The rising global burden of cancer has driven considerable efforts into the research and development of effective anti-cancer agents. Fortunately, with impressive advances in transcriptome profiling technology, the Connectivity Map (CMap) database has emerged as a promising and powerful drug repurposing approach. It provides an important platform for systematically discovering of the associations among genes, small-molecule compounds and diseases, and elucidating the mechanism of action of drug, contributing toward efficient anti-cancer pharmacotherapy. Moreover, CMap-based computational drug repurposing is gaining attention because of its potential to overcome the bottleneck constraints faced by traditional drug discovery in terms of cost, time and risk. Herein, we provide a comprehensive review of the applications of drug repurposing for anti-cancer drug discovery and summarize approaches for computational drug repurposing. We focus on the principle of the CMap database and novel CMap-based software/algorithms as well as their progress achieved for drug repurposing in the field of oncotherapy. This article is expected to illuminate the emerging potential of CMap in discovering effective anti-cancer drugs, thereby promoting efficient healthcare for cancer patients.

Alternative splicing (AS) is a key transcriptional regulation pathway. Recent studies have shown that AS events are associated with the occurrence of complex diseases. Various computational approaches have been developed for the detection of disease-associated AS events. In this review, we first describe the metrics used for quantitative characterization of AS events. Second, we review and discuss the three types of methods for detecting disease-associated splicing events, which are differential splicing analysis, aberrant splicing detection and splicing-related network analysis. Third, to further exploit the genetic mechanism of disease-associated AS events, we describe the methods for detecting genetic variants that potentially regulate splicing. For each type of methods, we conducted experimental comparison to illustrate their performance. Finally, we discuss the limitations of these methods and point out potential ways to address them. We anticipate that this review provides a systematic understanding of computational approaches for the analysis of disease-associated splicing.

Identifying disease-gene associations is a fundamental and critical biomedical task towards understanding molecular mechanisms, the diagnosis and treatment of diseases. It is time-consuming and expensive to experimentally verify causal links between diseases and genes. Recently, deep learning methods have achieved tremendous success in identifying candidate genes for genetic diseases. The gene prediction problem can be modeled as a link prediction problem based on the features of nodes and edges of the gene-disease graph. However, most existing researches either build homogeneous networks based on one single data source or heterogeneous networks based on multi-source data, and artificially define meta-paths, so as to learn the network representation of diseases and genes. The former cannot make use of abundant multi-source heterogeneous information, while the latter needs domain knowledge and experience when defining meta-paths, and the accuracy of the model largely depends on the definition of meta-paths. To address the aforementioned challenges above bottlenecks, we propose an end-to-end disease-gene association prediction model with parallel graph transformer network (DGP-PGTN), which deeply integrates the heterogeneous information of diseases, genes, ontologies and phenotypes. DGP-PGTN can automatically and comprehensively capture the multiple latent interactions between diseases and genes, discover the causal relationship between them and is fully interpretable at the same time. We conduct comprehensive experiments and show that DGP-PGTN outperforms the state-of-the-art methods significantly on the task of disease-gene association prediction. Furthermore, DGP-PGTN can automatically learn the implicit relationship between diseases and genes without manually defining meta paths.

The specificity of a T-cell receptor (TCR) repertoire determines personalized immune capacity. Existing methods have modeled the qualitative aspects of TCR specificity, while the quantitative aspects remained unaddressed. We developed a package, TCRanno, to quantify the specificity of TCR repertoires. We created deep-learning-based, epitope-aware vector embeddings to infer individual TCR specificity. Then we aggregated clonotype frequencies of TCRs to obtain a quantitative profile of repertoire specificity at epitope, antigen and organism levels. Applying TCRanno to 4195 TCR repertoires revealed quantitative changes in repertoire specificity upon infections, autoimmunity and cancers. Specifically, TCRanno found cytomegalovirus-specific TCRs in seronegative healthy individuals, supporting the possibility of abortive infections. TCRanno discovered age-accumulated fraction of severe acute respiratory syndrome coronavirus 2 specific TCRs in pre-pandemic samples, which may explain the aggressive symptoms and age-related severity of coronavirus disease 2019. TCRanno also identified the encounter of Hepatitis B antigens as a potential trigger of systemic lupus erythematosus. TCRanno annotations showed capability in distinguishing TCR repertoires of healthy and cancers including melanoma, lung and breast cancers. TCRanno also demonstrated usefulness to single-cell TCRseq+gene expression data analyses by isolating T-cells with the specificity of interest.

The critical first step in Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-associated (CRISPR-Cas) protein-mediated gene editing is recognizing a preferred protospacer adjacent motif (PAM) on target DNAs by the protein's PAM-interacting amino acids (PIAAs). Thus, accurate computational modeling of PAM recognition is useful in assisting CRISPR-Cas engineering to relax or tighten PAM requirements for subsequent applications. Here, we describe a universal computational protein design framework (UniDesign) for designing protein-nucleic acid interactions. As a proof of concept, we applied UniDesign to decode the PAM-PIAA interactions for eight Cas9 and two Cas12a proteins. We show that, given native PIAAs, the UniDesign-predicted PAMs are largely identical to the natural PAMs of all Cas proteins. In turn, given natural PAMs, the computationally redesigned PIAA residues largely recapitulated the native PIAAs (74% and 86% in terms of identity and similarity, respectively). These results demonstrate that UniDesign faithfully captures the mutual preference between natural PAMs and native PIAAs, suggesting it is a useful tool for engineering CRISPR-Cas and other nucleic acid-interacting proteins. UniDesign is open-sourced at https://github.com/tommyhuangthu/UniDesign.

Circular RNAs (circRNAs) are single-stranded and covalently closed non-coding RNA molecules originated from RNA splicing. Their functions include regulatory potential over other RNA species, such as microRNAs, messenger RNAs and RNA binding proteins. For circRNA identification, several algorithms are available and can be classified in two major types: pseudo-reference-based and split-alignment-based approaches. In general, the data generated from circRNA transcriptome initiatives is deposited on public specific databases, which provide a large amount of information on different species and functional annotations. In this review, we describe the main computational resources for the identification and characterization of circRNAs, covering the algorithms and predictive tools to evaluate its potential role in a particular transcriptomics project, including the public repositories containing relevant data and information for circRNAs, recapitulating their characteristics, reliability and amount of data reported.

Transcription factors (TFs) are key regulatory proteins that control the transcriptional rate of cells by binding short DNA sequences called transcription factor binding sites (TFBS) or motifs. Identifying and characterizing TFBS is fundamental to understanding the regulatory mechanisms governing the transcriptional state of cells. During the last decades, several experimental methods have been developed to recover DNA sequences containing TFBS. In parallel, computational methods have been proposed to discover and identify TFBS motifs based on these DNA sequences. This is one of the most widely investigated problems in bioinformatics and is referred to as the motif discovery problem. In this manuscript, we review classical and novel experimental and computational methods developed to discover and characterize TFBS motifs in DNA sequences, highlighting their advantages and drawbacks. We also discuss open challenges and future perspectives that could fill the remaining gaps in the field.

Since the 1980s, dozens of computational methods have addressed the problem of predicting RNA secondary structure. Among them are those that follow standard optimization approaches and, more recently, machine learning (ML) algorithms. The former were repeatedly benchmarked on various datasets. The latter, on the other hand, have not yet undergone extensive analysis that could suggest to the user which algorithm best fits the problem to be solved. In this review, we compare 15 methods that predict the secondary structure of RNA, of which 6 are based on deep learning (DL), 3 on shallow learning (SL) and 6 control methods on non-ML approaches. We discuss the ML strategies implemented and perform three experiments in which we evaluate the prediction of (I) representatives of the RNA equivalence classes, (II) selected Rfam sequences and (III) RNAs from new Rfam families. We show that DL-based algorithms (such as SPOT-RNA and UFold) can outperform SL and traditional methods if the data distribution is similar in the training and testing set. However, when predicting 2D structures for new RNA families, the advantage of DL is no longer clear, and its performance is inferior or equal to that of SL and non-ML methods.

BACKGROUND: Ubiquitous presence of short extrachromosomal circular DNAs (eccDNAs) in eukaryotic cells has perplexed generations of biologists. Their widespread origins in the genome lacking apparent specificity led some studies to conclude their formation as random or near-random. Despite this, the search for specific formation of short eccDNA continues with a recent surge of interest in biomarker development. RESULTS: To shed new light on the conflicting views on short eccDNAs' randomness, here we present DeepCircle, a bioinformatics framework incorporating convolution- and attention-based neural networks to assess their predictability. Short human eccDNAs from different datasets indeed have low similarity in genomic locations, but DeepCircle successfully learned shared DNA sequence features to make accurate cross-datasets predictions (accuracy: convolution-based models: 79.65 ± 4.7%, attention-based models: 83.31 ± 4.18%). CONCLUSIONS: The excellent performance of our models shows that the intrinsic predictability of eccDNAs is encoded in the sequences across tissue origins. Our work demonstrates how the perceived lack of specificity in genomics data can be re-assessed by deep learning models to uncover unexpected similarity.

MOTIVATION: Viruses have coevolved with their hosts for over millions of years and learned to escape the host's immune system. Although not all genetic changes in viruses are deleterious, some significant mutations lead to the escape of neutralizing antibodies and weaken the immune system, which increases infectivity and transmissibility, thereby impeding the development of antiviral drugs or vaccines. Accurate and reliable identification of viral escape mutational sequences could be a good indicator for therapeutic design. We developed a computational model that recognizes significant mutational sequences based on escape feature identification using natural language processing along with prior knowledge of experimentally validated escape mutants. RESULTS: Our machine learning-based computational approach can recognize the significant spike protein sequences of severe acute respiratory syndrome coronavirus 2 using sequence data alone. This modelling approach can be applied to other viruses, such as influenza, monkeypox and HIV using knowledge of escape mutants and relevant protein sequence datasets. AVAILABILITY: Complete source code and pre-trained models for escape prediction of severe acute respiratory syndrome coronavirus 2 protein sequences are available on Github at https://github.com/PremSinghBist/Sars-CoV-2-Escape-Model.git. The dataset is deposited to Zenodo at: doi: 10.5281/zenodo.7142638. The Python scripts are easy to run and customize as needed. CONTACT: premsing212@jbnu.ac.kr.

The ability to identify B-cell epitopes is an essential step in vaccine design, immunodiagnostic tests and antibody production. Several computational approaches have been proposed to identify, from an antigen protein or peptide sequence, which residues are more likely to be part of an epitope, but have limited performance on relatively homogeneous data sets and lack interpretability, limiting biological insights that could otherwise be obtained. To address these limitations, we have developed epitope1D, an explainable machine learning method capable of accurately identifying linear B-cell epitopes, leveraging two new descriptors: a graph-based signature representation of protein sequences, based on our well-established Cutoff Scanning Matrix algorithm and Organism Ontology information. Our model achieved Areas Under the ROC curve of up to 0.935 on cross-validation and blind tests, demonstrating robust performance. A comprehensive comparison to alternative methods using distinct benchmark data sets was also employed, with our model outperforming state-of-the-art tools. epitope1D represents not only a significant advance in predictive performance, but also allows biologically meaningful features to be combined and used for model interpretation. epitope1D has been made available as a user-friendly web server interface and application programming interface at https://biosig.lab.uq.edu.au/epitope1d/.

Network pharmacology is an emerging area of systematic drug research that attempts to understand drug actions and interactions with multiple targets. Network pharmacology has changed the paradigm from 'one-target one-drug' to highly potent 'multi-target drug'. Despite that, this synergistic approach is currently facing many challenges particularly mining effective information such as drug targets, mechanism of action, and drug and organism interaction from massive, heterogeneous data. To overcome bottlenecks in multi-target drug discovery, computational algorithms are highly welcomed by scientific community. Machine learning (ML) and especially its subfield deep learning (DL) have seen impressive advances. Techniques developed within these fields are now able to analyze and learn from huge amounts of data in disparate formats. In terms of network pharmacology, ML can improve discovery and decision making from big data. Opportunities to apply ML occur in all stages of network pharmacology research. Examples include screening of biologically active small molecules, target identification, metabolic pathways identification, protein-protein interaction network analysis, hub gene analysis and finding binding affinity between compounds and target proteins. This review summarizes the premier algorithmic concepts of ML in network pharmacology and forecasts future opportunities, potential applications as well as several remaining challenges of implementing ML in network pharmacology. To our knowledge, this study provides the first comprehensive assessment of ML approaches in network pharmacology, and we hope that it encourages additional efforts toward the development and acceptance of network pharmacology in the pharmaceutical industry.

The analysis of super-enhancers (SEs) has recently attracted attention in elucidating the molecular mechanisms of cancer and other diseases. SEs are genomic structures that strongly induce gene expression and have been reported to contribute to the overexpression of oncogenes. Because the analysis of SEs and integrated analysis with other data are performed using large amounts of genome-wide data, artificial intelligence technology, with machine learning at its core, has recently begun to be utilized. In promoting precision medicine, it is important to consider information from SEs in addition to genomic data; therefore, machine learning technology is expected to be introduced appropriately in terms of building a robust analysis platform with a high generalization performance. In this review, we explain the history and principles of SE, and the results of SE analysis using state-of-the-art machine learning and integrated analysis with other data are presented to provide a comprehensive understanding of the current status of SE analysis in the field of medical biology. Additionally, we compared the accuracy between existing machine learning methods on the benchmark dataset and attempted to explore the kind of data preprocessing and integration work needed to make the existing algorithms work on the benchmark dataset. Furthermore, we discuss the issues and future directions of current SE analysis.

The design of enzyme catalytic stability is of great significance in medicine and industry. However, traditional methods are time-consuming and costly. Hence, a growing number of complementary computational tools have been developed, e.g. ESMFold, AlphaFold2, Rosetta, RosettaFold, FireProt, ProteinMPNN. They are proposed for algorithm-driven and data-driven enzyme design through artificial intelligence (AI) algorithms including natural language processing, machine learning, deep learning, variational autoencoder/generative adversarial network, message passing neural network (MPNN). In addition, the challenges of design of enzyme catalytic stability include insufficient structured data, large sequence search space, inaccurate quantitative prediction, low efficiency in experimental validation and a cumbersome design process. The first principle of the enzyme catalytic stability design is to treat amino acids as the basic element. By designing the sequence of an enzyme, the flexibility and stability of the structure are adjusted, thus controlling the catalytic stability of the enzyme in a specific industrial environment or in an organism. Common indicators of design goals include the change in denaturation energy (ΔΔG), melting temperature (ΔTm), optimal temperature (Topt), optimal pH (pHopt), etc. In this review, we summarized and evaluated the enzyme design in catalytic stability by AI in terms of mechanism, strategy, data, labeling, coding, prediction, testing, unit, integration and prospect.

The expanding field of epitranscriptomics might rival the epigenome in the diversity of biological processes impacted. In recent years, the development of new high-throughput experimental and computational techniques has been a key driving force in discovering the properties of RNA modifications. Machine learning applications, such as for classification, clustering or de novo identification, have been critical in these advances. Nonetheless, various challenges remain before the full potential of machine learning for epitranscriptomics can be leveraged. In this review, we provide a comprehensive survey of machine learning methods to detect RNA modifications using diverse input data sources. We describe strategies to train and test machine learning methods and to encode and interpret features that are relevant for epitranscriptomics. Finally, we identify some of the current challenges and open questions about RNA modification analysis, including the ambiguity in predicting RNA modifications in transcript isoforms or in single nucleotides, or the lack of complete ground truth sets to test RNA modifications. We believe this review will inspire and benefit the rapidly developing field of epitranscriptomics in addressing the current limitations through the effective use of machine learning.

RNA methylation has emerged recently as an active research domain to study post-transcriptional alteration in gene expression regulation. Various types of RNA methylation, including N6-methyladenosine (m6A), are involved in human disease development. As a newly developed sequencing biotechnology to quantify the m6A level on a transcriptome-wide scale, MeRIP-seq expands RNA epigenetics study in both basic and clinical applications, with an upward trend. One of the fundamental questions in RNA methylation data analysis is to identify the Differentially Methylated Regions (DMRs), by contrasting cases and controls. Multiple statistical approaches have been recently developed for DMR detection, but there is a lack of a comprehensive evaluation for these analytical methods. Here, we thoroughly assess all eight existing methods for DMR calling, using both synthetic and real data. Our simulation adopts a Gamma-Poisson model and logit linear framework, and accommodates various sample sizes and DMR proportions for benchmarking. For all methods, low sensitivities are observed among regions with low input levels, but they can be drastically boosted by an increase in sample size. TRESS and exomePeak2 perform the best using metrics of detection precision, FDR, type I error control and runtime, though hampered by low sensitivity. DRME and exomePeak obtain high sensitivities, at the expense of inflated FDR and type I error. Analyses on three real datasets suggest differential preference on identified DMR length and uniquely discovered regions, between these methods.

In the era of constantly increasing amounts of the available protein data, a relevant and interpretable visualization becomes crucial, especially for tasks requiring human expertise. Poincaré disk projection has previously demonstrated its important efficiency for visualization of biological data such as single-cell RNAseq data. Here, we develop a new method PoincaréMSA for visual representation of complex relationships between protein sequences based on Poincaré maps embedding. We demonstrate its efficiency and potential for visualization of protein family topology as well as evolutionary and functional annotation of uncharacterized sequences. PoincaréMSA is implemented in open source Python code with available interactive Google Colab notebooks as described at https://www.dsimb.inserm.fr/POINCARE_MSA.

The off-target effect occurring in the CRISPR-Cas9 system has been a challenging problem for the practical application of this gene editing technology. In recent years, various prediction models have been proposed to predict potential off-target activities. However, most of the existing prediction methods do not fully exploit guide RNA (gRNA) and DNA sequence pair information effectively. In addition, available prediction methods usually ignore the noise effect in original off-target datasets. To address these issues, we design a novel coding scheme, which considers the key features of mismatch type, mismatch location and the gRNA-DNA sequence pair information. Furthermore, a transformer-based anti-noise model called CrisprDNT is developed to solve the noise problem that exists in the off-target data. Experimental results of eight existing datasets demonstrate that the method with the inclusion of the anti-noise loss functions is superior to available state-of-the-art prediction methods. CrisprDNT is available at https://github.com/gzrgzx/CrisprDNT.

Cooperative driver pathways discovery helps researchers to study the pathogenesis of cancer. However, most discovery methods mainly focus on genomics data, and neglect the known pathway information and other related multi-omics data; thus they cannot faithfully decipher the carcinogenic process. We propose CDPMiner (Cooperative Driver Pathways Miner) to discover cooperative driver pathways by multiplex network embedding, which can jointly model relational and attribute information of multi-type molecules. CDPMiner first uses the pathway topology to quantify the weight of genes in different pathways, and optimizes the relations between genes and pathways. Then it constructs an attributed multiplex network consisting of micro RNAs, long noncoding RNAs, genes and pathways, embeds the network through deep joint matrix factorization to mine more essential information for pathway-level analysis and reconstructs the pathway interaction network. Finally, CDPMiner leverages the reconstructed network and mutation data to define the driver weight between pathways to discover cooperative driver pathways. Experimental results on Breast invasive carcinoma and Stomach adenocarcinoma datasets show that CDPMiner can effectively fuse multi-omics data to discover more driver pathways, which indeed cooperatively trigger cancers and are valuable for carcinogenesis analysis. Ablation study justifies CDPMiner for a more comprehensive analysis of cancer by fusing multi-omics data.

With the rapid development of human intestinal microbiology and diverse microbiome-related studies and investigations, a large amount of data have been generated and accumulated. Meanwhile, different computational and bioinformatics models have been developed for pattern recognition and knowledge discovery using these data. Given the heterogeneity of these resources and models, we aimed to provide a landscape of the data resources, a comparison of the computational models and a summary of the translational informatics applied to microbiota data. We first review the existing databases, knowledge bases, knowledge graphs and standardizations of microbiome data. Then, the high-throughput sequencing techniques for the microbiome and the informatics tools for their analyses are compared. Finally, translational informatics for the microbiome, including biomarker discovery, personalized treatment and smart healthcare for complex diseases, are discussed.

Advancing spatially resolved transcriptomics (ST) technologies help biologists comprehensively understand organ function and tissue microenvironment. Accurate spatial domain identification is the foundation for delineating genome heterogeneity and cellular interaction. Motivated by this perspective, a graph deep learning (GDL) based spatial clustering approach is constructed in this paper. First, the deep graph infomax module embedded with residual gated graph convolutional neural network is leveraged to address the gene expression profiles and spatial positions in ST. Then, the Bayesian Gaussian mixture model is applied to handle the latent embeddings to generate spatial domains. Designed experiments certify that the presented method is superior to other state-of-the-art GDL-enabled techniques on multiple ST datasets. The codes and dataset used in this manuscript are summarized at https://github.com/narutoten520/SCGDL.

Recovering high-quality metagenome-assembled genomes (HQ-MAGs) is critical for exploring microbial compositions and microbe-phenotype associations. However, multiple sequencing platforms and computational tools for this purpose may confuse researchers and thus call for extensive evaluation. Here, we systematically evaluated a total of 40 combinations of popular computational tools and sequencing platforms (i.e. strategies), involving eight assemblers, eight metagenomic binners and four sequencing technologies, including short-, long-read and metaHiC sequencing. We identified the best tools for the individual tasks (e.g. the assembly and binning) and combinations (e.g. generating more HQ-MAGs) depending on the availability of the sequencing data. We found that the combination of the hybrid assemblies and metaHiC-based binning performed best, followed by the hybrid and long-read assemblies. More importantly, both long-read and metaHiC sequencings link more mobile elements and antibiotic resistance genes to bacterial hosts and improve the quality of public human gut reference genomes with 32% (34/105) HQ-MAGs that were either of better quality than those in the Unified Human Gastrointestinal Genome catalog version 2 or novel.

Gaussian graphical model is a strong tool for identifying interactions from metabolomics data based on conditional correlation. However, data may be collected from different stages or subgroups of subjects with heterogeneity or hierarchical structure. There are different integrating strategies of graphical models for multi-group data proposed by data scientists. It is challenging to select the methods for metabolism data analysis. This study aimed to evaluate the performance of several different integrating graphical models for multi-group data and provide support for the choice of strategy for similar characteristic data. We compared the performance of seven methods in estimating graph structures through simulation study. We also applied all the methods in breast cancer metabolomics data grouped by stages to illustrate the real data application. The method of Shaddox et al. achieved the highest average area under the receiver operating characteristic curve and area under the precision-recall curve across most scenarios, and it was the only approach with all indicators ranked at the top. Nevertheless, it also cost the most time in all settings. Stochastic search structure learning tends to result in estimates that focus on the precision of identified edges, while BEAM, hierarchical Bayesian approach and birth-death Markov chain Monte Carlo may identify more potential edges. In the real metabolomics data analysis from three stages of breast cancer patients, results were in line with that in simulation study.

The dynamical properties of many complex physical and biological systems can be quantified from the energy landscape theory. Previous approaches focused on estimating the transition rate from landscape reconstruction based on data. However, for general non-equilibrium systems (such as gene regulatory systems), both the energy landscape and the probability flux are important to determine the transition rate between attractors. In this work, we proposed a data-driven approach to estimate non-equilibrium transition rate, which combines the kernel density estimation and non-equilibrium transition rate theory. Our approach shows superior performance in estimating transition rate from data, compared with previous methods, due to the introduction of a nonparametric density estimation method and the new saddle point by considering the effects of flux. We demonstrate the practical validity of our approach by applying it to a simplified cell fate decision model and a high-dimensional stem cell differentiation model. Our approach can be applied to other biological and physical systems.

CRISPR/Cas9 (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-associated protein 9) is a popular and effective two-component technology used for targeted genetic manipulation. It is currently the most versatile and accurate method of gene and genome editing, which benefits from a large variety of practical applications. For example, in biomedicine, it has been used in research related to cancer, virus infections, pathogen detection, and genetic diseases. Current CRISPR/Cas9 research is based on data-driven models for on- and off-target prediction as a cleavage may occur at non-target sequence locations. Nowadays, conventional machine learning and deep learning methods are applied on a regular basis to accurately predict on-target knockout efficacy and off-target profile of given single-guide RNAs (sgRNAs). In this paper, we present an overview and a comparative analysis of traditional machine learning and deep learning models used in CRISPR/Cas9. We highlight the key research challenges and directions associated with target activity prediction. We discuss recent advances in the sgRNA-DNA sequence encoding used in state-of-the-art on- and off-target prediction models. Furthermore, we present the most popular deep learning neural network architectures used in CRISPR/Cas9 prediction models. Finally, we summarize the existing challenges and discuss possible future investigations in the field of on- and off-target prediction. Our paper provides valuable support for academic and industrial researchers interested in the application of machine learning methods in the field of CRISPR/Cas9 genome editing.

In genome assembly, scaffolding can obtain more complete and continuous scaffolds. Current scaffolding methods usually adopt one type of read to construct a scaffold graph and then orient and order contigs. However, scaffolding with the strengths of two or more types of reads seems to be a better solution to some tricky problems. Combining the advantages of different types of data is significant for scaffolding. Here, a hybrid scaffolding method (SLHSD) is present that simultaneously leverages the precision of short reads and the length advantage of long reads. Building an optimal scaffold graph is an important foundation for getting scaffolds. SLHSD uses a new algorithm that combines long and short read alignment information to determine whether to add an edge and how to calculate the edge weight in a scaffold graph. In addition, SLHSD develops a strategy to ensure that edges with high confidence can be added to the graph with priority. Then, a linear programming model is used to detect and remove remaining false edges in the graph. We compared SLHSD with other scaffolding methods on five datasets. Experimental results show that SLHSD outperforms other methods. The open-source code of SLHSD is available at https://github.com/luojunwei/SLHSD.

Imaging mass spectrometry (IMS) is one of the powerful tools in spatial metabolomics for obtaining metabolite data and probing the internal microenvironment of organisms. It has dramatically advanced the understanding of the structure of biological tissues and the drug treatment of diseases. However, the complexity of IMS data hinders the further acquisition of biomarkers and the study of certain specific activities of organisms. To this end, we introduce an artificial intelligence tool, SmartGate, to enable automatic peak selection and spatial structure identification in an iterative manner. SmartGate selects discriminative m/z features from the previous iteration by differential analysis and employs a graph attention autoencoder model to perform spatial clustering for tissue segmentation using the selected features. We applied SmartGate to diverse IMS data at multicellular or subcellular spatial resolutions and compared it with four competing methods to demonstrate its effectiveness. SmartGate can significantly improve the accuracy of spatial segmentation and identify biomarker metabolites based on tissue structure-guided differential analysis. For multiple consecutive IMS data, SmartGate can effectively identify structures with spatial heterogeneity by introducing three-dimensional spatial neighbor information.

Rapid and accurate prediction of drug-target affinity can accelerate and improve the drug discovery process. Recent studies show that deep learning models may have the potential to provide fast and accurate drug-target affinity prediction. However, the existing deep learning models still have their own disadvantages that make it difficult to complete the task satisfactorily. Complex-based models rely heavily on the time-consuming docking process, and complex-free models lacks interpretability. In this study, we introduced a novel knowledge-distillation insights drug-target affinity prediction model with feature fusion inputs to make fast, accurate and explainable predictions. We benchmarked the model on public affinity prediction and virtual screening dataset. The results show that it outperformed previous state-of-the-art models and achieved comparable performance to previous complex-based models. Finally, we study the interpretability of this model through visualization and find it can provide meaningful explanations for pairwise interaction. We believe this model can further improve the drug-target affinity prediction for its higher accuracy and reliable interpretability.

Major histocompatibility complex (MHC) class II molecules play a pivotal role in antigen presentation and CD4+ T cell response. Accurate prediction of the immunogenicity of MHC class II-associated antigens is critical for vaccine design and cancer immunotherapies. However, current computational methods are limited by insufficient training data and algorithmic constraints, and the rules that govern which peptides are truly recognized by existing T cell receptors remain poorly understood. Here, we build a transfer learning-based, long short-term memory model named 'TLimmuno2' to predict whether epitope-MHC class II complex can elicit T cell response. Through leveraging binding affinity data, TLimmuno2 shows superior performance compared with existing models on independent validation datasets. TLimmuno2 can find real immunogenic neoantigen in real-world cancer immunotherapy data. The identification of significant MHC class II neoantigen-mediated immunoediting signal in the cancer genome atlas pan-cancer dataset further suggests the robustness of TLimmuno2 in identifying really immunogenic neoantigens that are undergoing negative selection during cancer evolution. Overall, TLimmuno2 is a powerful tool for the immunogenicity prediction of MHC class II presented epitopes and could promote the development of personalized immunotherapies.

AlphaFold2 achieved a breakthrough in protein structure prediction through the end-to-end deep learning method, which can predict nearly all single-domain proteins at experimental resolution. However, the prediction accuracy of full-chain proteins is generally lower than that of single-domain proteins because of the incorrect interactions between domains. In this work, we develop an inter-domain distance prediction method, named DeepIDDP. In DeepIDDP, we design a neural network with attention mechanisms, where two new inter-domain features are used to enhance the ability to capture the interactions between domains. Furthermore, we propose a data enhancement strategy termed DPMSA, which is employed to deal with the absence of co-evolutionary information on targets. We integrate DeepIDDP into our previously developed domain assembly method SADA, termed SADA-DeepIDDP. Tested on a given multi-domain benchmark dataset, the accuracy of SADA-DeepIDDP inter-domain distance prediction is 11.3% and 21.6% higher than trRosettaX and trRosetta, respectively. The accuracy of the domain assembly model is 2.5% higher than that of SADA. Meanwhile, we reassemble 68 human multi-domain protein models with TM-score ≤ 0.80 from the AlphaFold protein structure database, where the average TM-score is improved by 11.8% after the reassembly by our method. The online server is at http://zhanglab-bioinf.com/DeepIDDP/.

Deciphering cell-type-specific 3D structures of chromatin is challenging. Here, we present InferLoop, a novel method for inferring the strength of chromatin interaction using single-cell chromatin accessibility data. The workflow of InferLoop is, first, to conduct signal enhancement by grouping nearby cells into bins, and then, for each bin, leverage accessibility signals for loop signals using a newly constructed metric that is similar to the perturbation of the Pearson correlation coefficient. In this study, we have described three application scenarios of InferLoop, including the inference of cell-type-specific loop signals, the prediction of gene expression levels and the interpretation of intergenic loci. The effectiveness and superiority of InferLoop over other methods in those three scenarios are rigorously validated by using the single-cell 3D genome structure data of human brain cortex and human blood, the single-cell multi-omics data of human blood and mouse brain cortex, and the intergenic loci in the GWAS Catalog database as well as the GTEx database, respectively. In addition, InferLoop can be applied to predict loop signals of individual spots using the spatial chromatin accessibility data of mouse embryo. InferLoop is available at https://github.com/jumphone/inferloop.

Doublets formed during single-cell RNA sequencing (scRNA-seq) severely affect downstream studies, such as differentially expressed gene analysis and cell trajectory inference, and limit the cellular throughput of scRNA-seq. Several doublet detection algorithms are currently available, but their generalization performance could be further improved due to the lack of effective feature-embedding strategies with suitable model architectures. Therefore, SoCube, a novel deep learning algorithm, was developed to precisely detect doublets in various types of scRNA-seq data. SoCube (i) proposed a novel 3D composite feature-embedding strategy that embedded latent gene information and (ii) constructed a multikernel, multichannel CNN-ensembled architecture in conjunction with the feature-embedding strategy. With its excellent performance on benchmark evaluation and several downstream tasks, it is expected to be a powerful algorithm to detect and remove doublets in scRNA-seq data. SoCube is freely provided as an end-to-end tool on the Python official package site PyPi (https://pypi.org/project/socube/) and open-source on GitHub (https://github.com/idrblab/socube/).

Combination therapy is widely used to treat complex diseases, particularly in patients who respond poorly to monotherapy. For example, compared with the use of a single drug, drug combinations can reduce drug resistance and improve the efficacy of cancer treatment. Thus, it is vital for researchers and society to help develop effective combination therapies through clinical trials. However, high-throughput synergistic drug combination screening remains challenging and expensive in the large combinational space, where an array of compounds are used. To solve this problem, various computational approaches have been proposed to effectively identify drug combinations by utilizing drug-related biomedical information. In this study, considering the implications of various types of neighbor information of drug entities, we propose a novel end-to-end Knowledge Graph Attention Network to predict drug synergy (KGANSynergy), which utilizes neighbor information of known drugs/cell lines effectively. KGANSynergy uses knowledge graph (KG) hierarchical propagation to find multi-source neighbor nodes for drugs and cell lines. The knowledge graph attention network is designed to distinguish the importance of neighbors in a KG through a multi-attention mechanism and then aggregate the entity's neighbor node information to enrich the entity. Finally, the learned drug and cell line embeddings can be utilized to predict the synergy of drug combinations. Experiments demonstrated that our method outperformed several other competing methods, indicating that our method is effective in identifying drug combinations.

In cryogenic electron microscopy (cryo-EM) single particle analysis (SPA), high-resolution three-dimensional structures of biological macromolecules are determined by iteratively aligning and averaging a large number of two-dimensional projections of molecules. Since the correlation measures are sensitive to the signal-to-noise ratio, various parameter estimation steps in SPA will be disturbed by the high-intensity noise in cryo-EM. However, denoising algorithms tend to damage high frequencies and suppress mid- and high-frequency contrast of micrographs, which exactly the precise parameter estimation relies on, therefore, limiting their application in SPA. In this study, we suggest combining a cryo-EM image processing pipeline with denoising and maximizing the signal's contribution in various parameter estimation steps. To solve the inherent flaws of denoising algorithms, we design an algorithm named MScale to correct the amplitude distortion caused by denoising and propose a new orientation determination strategy to compensate for the high-frequency loss. In the experiments on several real datasets, the denoised particles are successfully applied in the class assignment estimation and orientation determination tasks, ultimately enhancing the quality of biomacromolecule reconstruction. The case study on classification indicates that our strategy not only improves the resolution of difficult classes (up to 5 Å) but also resolves an additional class. In the case study on orientation determination, our strategy improves the resolution of the final reconstructed density map by 0.34 Å compared with conventional strategy. The code is available at https://github.com/zhanghui186/Mscale.

MOTIVATION: Interpretability has become a necessary feature for machine learning models deployed in critical scenarios, e.g. legal system, healthcare. In these situations, algorithmic decisions may have (potentially negative) long-lasting effects on the end-user affected by the decision. While deep learning models achieve impressive results, they often function as a black-box. Inspired by linear models, we propose a novel class of structurally constrained deep neural networks, which we call FLAN (Feature-wise Latent Additive Networks). Crucially, FLANs process each input feature separately, computing for each of them a representation in a common latent space. These feature-wise latent representations are then simply summed, and the aggregated representation is used for the prediction. These feature-wise representations allow a user to estimate the effect of each individual feature independently from the others, similarly to the way linear models are interpreted. RESULTS: We demonstrate FLAN on a series of benchmark datasets in different biological domains. Our experiments show that FLAN achieves good performances even in complex datasets (e.g. TCR-epitope binding prediction), despite the structural constraint we imposed. On the other hand, this constraint enables us to interpret FLAN by deciphering its decision process, as well as obtaining biological insights (e.g. by identifying the marker genes of different cell populations). In supplementary experiments, we show similar performances also on non-biological datasets. CODE AND DATA AVAILABILITY: Code and example data are available at https://github.com/phineasng/flan_bio.

Single-cell CRISPR screens have been widely used to investigate gene regulatory circuits in diverse biological systems. The recent development of single-cell CRISPR screens has enabled multimodal profiling of perturbed cells with both gene expression, chromatin accessibility and protein levels. However, current methods cannot meet the analysis requirements of different types of data and have limited functions. Here, we introduce Single-cell CRISPR screens data analysEs and perturbation modEling (SCREE) as a comprehensive and flexible pipeline to facilitate the analyses of various types of single-cell CRISPR screens data. SCREE performs read alignment, sgRNA assignment, quality control, clustering and visualization, perturbation enrichment evaluation, perturbation efficiency modeling, gene regulatory score calculation and functional analyses of perturbations for single-cell CRISPR screens with both RNA, ATAC and multimodal readout. SCREE is available at https://github.com/wanglabtongji/SCREE.

Peptide-major histocompatibility complex I (MHC I) binding affinity prediction is crucial for vaccine development, but existing methods face limitations such as small datasets, model overfitting due to excessive parameters and suboptimal performance. Here, we present STMHCPan (STAR-MHCPan), an open-source package based on the Star-Transformer model, for MHC I binding peptide prediction. Our approach introduces an attention mechanism to improve the deep learning network architecture and performance in antigen prediction. Compared with classical deep learning algorithms, STMHCPan exhibits improved performance with fewer parameters in receptor affinity training. Furthermore, STMHCPan outperforms existing ligand benchmark datasets identified by mass spectrometry. It can also handle peptides of arbitrary length and is highly scalable for predicting T-cell responses. Our software is freely available for use, training and extension through Github (https://github.com/Luckysoutheast/STMHCPan.git).

Single-cell omics data are growing at an unprecedented rate, whereas effective integration of them remains challenging due to different sequencing methods, quality, and expression pattern of each omics data. In this study, we propose a universal framework for the integration of single-cell multi-omics data based on graph convolutional network (GCN-SC). Among the multiple single-cell data, GCN-SC usually selects one data with the largest number of cells as the reference and the rest as the query dataset. It utilizes mutual nearest neighbor algorithm to identify cell-pairs, which provide connections between cells both within and across the reference and query datasets. A GCN algorithm further takes the mixed graph constructed from these cell-pairs to adjust count matrices from the query datasets. Finally, dimension reduction is performed by using non-negative matrix factorization before visualization. By applying GCN-SC on six datasets, we show that GCN-SC can effectively integrate sequencing data from multiple single-cell sequencing technologies, species or different omics, which outperforms the state-of-the-art methods, including Seurat, LIGER, GLUER and Pamona.

MicroRNAs are small regulatory RNAs that decrease gene expression after transcription in various biological disciplines. In bioinformatics, identifying microRNAs and predicting their functionalities is critical. Finding motifs is one of the most well-known and important methods for identifying the functionalities of microRNAs. Several motif discovery techniques have been proposed, some of which rely on artificial intelligence-based techniques. However, in the case of few or no training data, their accuracy is low. In this research, we propose a new computational approach, called DiMo, for identifying motifs in microRNAs and generally macromolecules of small length. We employ word embedding techniques and deep learning models to improve the accuracy of motif discovery results. Also, we rely on transfer learning models to pre-train a model and use it in cases of a lack of (enough) training data. We compare our approach with five state-of-the-art works using three real-world datasets. DiMo outperforms the selected related works in terms of precision, recall, accuracy and f1-score.

Although many single-cell computational methods proposed use gene expression as input, recent studies show that replacing 'unstable' gene expression with 'stable' gene-gene associations can greatly improve the performance of downstream analysis. To obtain accurate gene-gene associations, conditional cell-specific network method (c-CSN) filters out the indirect associations of cell-specific network method (CSN) based on the conditional independence of statistics. However, when there are strong connections in networks, the c-CSN suffers from false negative problem in network construction. To overcome this problem, a new partial cell-specific network method (p-CSN) based on the partial independence of statistics is proposed in this paper, which eliminates the singularity of the c-CSN by implicitly including direct associations among estimated variables. Based on the p-CSN, single-cell network entropy (scNEntropy) is further proposed to quantify cell state. The superiorities of our method are verified on several datasets. (i) Compared with traditional gene regulatory network construction methods, the p-CSN constructs partial cell-specific networks, namely, one cell to one network. (ii) When there are strong connections in networks, the p-CSN reduces the false negative probability of the c-CSN. (iii) The input of more accurate gene-gene associations further optimizes the performance of downstream analyses. (iv) The scNEntropy effectively quantifies cell state and reconstructs cell pseudo-time.

MOTIVATION: Histones are the chief protein components of chromatin, and the chemical modifications on histones crucially influence the transcriptional state of related genes. Histone modifying enzyme (HME), responsible for adding or removing the chemical labels, has emerged as a very important class of drug target, with a few HME inhibitors launched as anti-cancerous drugs and tens of molecules under clinical trials. To accelerate the drug discovery process of HME inhibitors, machine learning-based predictive models have been developed to enrich the active molecules from vast chemical space. However, the number of compounds with known activity distributed largely unbalanced among different HMEs, particularly with many targets of less than a hundred active samples. In this case, it is difficult to build effective virtual screening models directly based on machine learning. RESULTS: To this end, we propose a new Meta-learning-based Histone Modifying Enzymes Inhibitor prediction method (MetaHMEI). Our proposed MetaHMEI first uses a self-supervised pre-training approach to obtain high-quality molecular substructure embeddings from a large unlabeled chemical dataset. Then, MetaHMEI exploits a Transformer-based encoder and meta-learning framework to build a prediction model. MetaHMEI allows the effective transfer of the prior knowledge learned from HMEs with sufficient samples to HMEs with a small number of samples, so the proposed model can produce accurate predictions for HMEs with limited data. Extensive experimental results on our collected and curated HMEs datasets show that MetaHMEI is better than other methods in the case of few-shot learning. Furthermore, we applied MetaHMEI in the virtual screening process of histone JMJD3 inhibitors and successfully obtained three small molecule inhibitors, further supporting the validity of our model.

Haplotype networks are graphs used to represent evolutionary relationships between a set of taxa and are characterized by intuitiveness in analyzing genealogical relationships of closely related genomes. We here propose a novel algorithm termed McAN that considers mutation spectrum history (mutations in ancestry haplotype should be contained in descendant haplotype), node size (corresponding to sample count for a given node) and sampling time when constructing haplotype network. We show that McAN is two orders of magnitude faster than state-of-the-art algorithms without losing accuracy, making it suitable for analysis of a large number of sequences. Based on our algorithm, we developed an online web server and offline tool for haplotype network construction, community lineage determination, and interactive network visualization. We demonstrate that McAN is highly suitable for analyzing and visualizing massive genomic data and is helpful to enhance the understanding of genome evolution. Availability: Source code is written in C/C++ and available at https://github.com/Theory-Lun/McAN and https://ngdc.cncb.ac.cn/biocode/tools/BT007301 under the MIT license. Web server is available at https://ngdc.cncb.ac.cn/bit/hapnet/. SARS-CoV-2 dataset are available at https://ngdc.cncb.ac.cn/ncov/. Contact: songshh@big.ac.cn (Song S), zhaowm@big.ac.cn (Zhao W), baoym@big.ac.cn (Bao Y), zhangzhang@big.ac.cn (Zhang Z), ybxue@big.ac.cn (Xue Y).

With the aim of analyzing large-sized multidimensional single-cell datasets, we are describing a method for Cosine-based Tanimoto similarity-refined graph for community detection using Leiden's algorithm (CosTaL). As a graph-based clustering method, CosTaL transforms the cells with high-dimensional features into a weighted k-nearest-neighbor (kNN) graph. The cells are represented by the vertices of the graph, while an edge between two vertices in the graph represents the close relatedness between the two cells. Specifically, CosTaL builds an exact kNN graph using cosine similarity and uses the Tanimoto coefficient as the refining strategy to re-weight the edges in order to improve the effectiveness of clustering. We demonstrate that CosTaL generally achieves equivalent or higher effectiveness scores on seven benchmark cytometry datasets and six single-cell RNA-sequencing datasets using six different evaluation metrics, compared with other state-of-the-art graph-based clustering methods, including PhenoGraph, Scanpy and PARC. As indicated by the combined evaluation metrics, Costal has high efficiency with small datasets and acceptable scalability for large datasets, which is beneficial for large-scale analysis.

The advances of single-cell transcriptomic technologies have led to increasing use of single-cell RNA sequencing (scRNA-seq) data in large-scale patient cohort studies. The resulting high-dimensional data can be summarized and incorporated into patient outcome prediction models in several ways; however, there is a pressing need to understand the impact of analytical decisions on such model quality. In this study, we evaluate the impact of analytical choices on model choices, ensemble learning strategies and integrate approaches on patient outcome prediction using five scRNA-seq COVID-19 datasets. First, we examine the difference in performance between using single-view feature space versus multi-view feature space. Next, we survey multiple learning platforms from classical machine learning to modern deep learning methods. Lastly, we compare different integration approaches when combining datasets is necessary. Through benchmarking such analytical combinations, our study highlights the power of ensemble learning, consistency among different learning methods and robustness to dataset normalization when using multiple datasets as the model input.

Immunofluorescence patterns of anti-nuclear antibodies (ANAs) on human epithelial cell (HEp-2) substrates are important biomarkers for the diagnosis of autoimmune diseases. There are growing clinical requirements for an automatic readout and classification of ANA immunofluorescence patterns for HEp-2 images following the taxonomy recommended by the International Consensus on Antinuclear Antibody Patterns (ICAP). In this study, a comprehensive collection of HEp-2 specimen images covering a broad range of ANA patterns was established and manually annotated by experienced laboratory experts. By utilizing a supervised learning methodology, an automatic immunofluorescence pattern classification framework for HEp-2 specimen images was developed. The framework consists of a module for HEp-2 cell detection and cell-level feature extraction, followed by an image-level classifier that is capable of recognizing all 14 classes of ANA immunofluorescence patterns as recommended by ICAP. Performance analysis indicated an accuracy of 92.05% on the validation dataset and 87% on an independent test dataset, which has surpassed the performance of human examiners on the same test dataset. The proposed framework is expected to contribute to the automatic ANA pattern recognition in clinical laboratories to facilitate efficient and precise diagnosis of autoimmune diseases.

Boolean models are a well-established framework to model developmental gene regulatory networks (DGRNs) for acquisition of cellular identities. During the reconstruction of Boolean DGRNs, even if the network structure is given, there is generally a large number of combinations of Boolean functions that will reproduce the different cell fates (biological attractors). Here we leverage the developmental landscape to enable model selection on such ensembles using the relative stability of the attractors. First we show that previously proposed measures of relative stability are strongly correlated and we stress the usefulness of the one that captures best the cell state transitions via the mean first passage time (MFPT) as it also allows the construction of a cellular lineage tree. A property of great computational importance is the insensitivity of the different stability measures to changes in noise intensities. That allows us to use stochastic approaches to estimate the MFPT and thereby scale up the computations to large networks. Given this methodology, we revisit different Boolean models of Arabidopsis thaliana root development, showing that a most recent one does not respect the biologically expected hierarchy of cell states based on relative stabilities. We therefore developed an iterative greedy algorithm that searches for models which satisfy the expected hierarchy of cell states and found that its application to the root development model yields many models that meet this expectation. Our methodology thus provides new tools that can enable reconstruction of more realistic and accurate Boolean models of DGRNs.

Intra-tumor heterogeneity (ITH) is one of the major confounding factors that result in cancer relapse, and deciphering ITH is essential for personalized therapy. Single-cell DNA sequencing (scDNA-seq) now enables profiling of single-cell copy number alterations (CNAs) and thus aids in high-resolution inference of ITH. Here, we introduce an integrated framework called rcCAE to accurately infer cell subpopulations and single-cell CNAs from scDNA-seq data. A convolutional autoencoder (CAE) is employed in rcCAE to learn latent representation of the cells as well as distill copy number information from noisy read counts data. This unsupervised representation learning via the CAE model makes it convenient to accurately cluster cells over the low-dimensional latent space, and detect single-cell CNAs from enhanced read counts data. Extensive performance evaluations on simulated datasets show that rcCAE outperforms the existing CNA calling methods, and is highly effective in inferring clonal architecture. Furthermore, evaluations of rcCAE on two real datasets demonstrate that it is able to provide a more refined clonal structure, of which some details are lost in clonal inference based on integer copy numbers.

Colorectal cancer (CRC) is one of the most common gastrointestinal malignancies. There are few recurrence risk signatures for CRC patients. Single-cell RNA-sequencing (scRNA-seq) provides a high-resolution platform for prognostic signature detection. However, scRNA-seq is not practical in large cohorts due to its high cost and most single-cell experiments lack clinical phenotype information. Few studies have been reported to use external bulk transcriptome with survival time to guide the detection of key cell subtypes in scRNA-seq data. We proposed scRankXMBD, a computational framework to prioritize prognostic-associated cell subpopulations based on within-cell relative expression orderings of gene pairs from single-cell transcriptomes. scRankXMBD achieves higher precision and concordance compared with five existing methods. Moreover, we developed single-cell gene pair signatures to predict recurrence risk for patients individually. Our work facilitates the application of the rank-based method in scRNA-seq data for prognostic biomarker discovery and precision oncology. scRankXMBD is available at https://github.com/xmuyulab/scRank-XMBD. (XMBD:Xiamen Big Data, a biomedical open software initiative in the National Institute for Data Science in Health and Medicine, Xiamen University, China.).

Single-cell RNA sequencing (scRNA-seq) is a revolutionary breakthrough that determines the precise gene expressions on individual cells and deciphers cell heterogeneity and subpopulations. However, scRNA-seq data are much noisier than traditional high-throughput RNA-seq data because of technical limitations, leading to many scRNA-seq data studies about dimensionality reduction and visualization remaining at the basic data-stacking stage. In this study, we propose an improved variational autoencoder model (termed DREAM) for dimensionality reduction and a visual analysis of scRNA-seq data. Here, DREAM combines the variational autoencoder and Gaussian mixture model for cell type identification, meanwhile explicitly solving 'dropout' events by introducing the zero-inflated layer to obtain the low-dimensional representation that describes the changes in the original scRNA-seq dataset. Benchmarking comparisons across nine scRNA-seq datasets show that DREAM outperforms four state-of-the-art methods on average. Moreover, we prove that DREAM can accurately capture the expression dynamics of human preimplantation embryonic development. DREAM is implemented in Python, freely available via the GitHub website, https://github.com/Crystal-JJ/DREAM.

Understanding the interactions between the biomolecules that govern cellular behaviors remains an emergent question in biology. Recent advances in single-cell technologies have enabled the simultaneous quantification of multiple biomolecules in the same cell, opening new avenues for understanding cellular complexity and heterogeneity. Still, the resulting multimodal single-cell datasets present unique challenges arising from the high dimensionality and multiple sources of acquisition noise. Computational methods able to match cells across different modalities offer an appealing alternative towards this goal. In this work, we propose MatchCLOT, a novel method for modality matching inspired by recent promising developments in contrastive learning and optimal transport. MatchCLOT uses contrastive learning to learn a common representation between two modalities and applies entropic optimal transport as an approximate maximum weight bipartite matching algorithm. Our model obtains state-of-the-art performance on two curated benchmarking datasets and an independent test dataset, improving the top scoring method by 26.1% while preserving the underlying biological structure of the multimodal data. Importantly, MatchCLOT offers high gains in computational time and memory that, in contrast to existing methods, allows it to scale well with the number of cells. As single-cell datasets become increasingly large, MatchCLOT offers an accurate and efficient solution to the problem of modality matching.

Exosomes cargo tumour-characterized biomolecules secreted from cancer cells and play a pivotal role in tumorigenesis and cancer progression, thus providing their potential for non-invasive cancer monitoring. Since cancer cell-derived exosomes are often mixed with those from healthy cells in liquid biopsy of tumour patients, accurately measuring the purity of tumour cell-derived exosomes is not only critical for the early detection but also essential for unbiased identification of diagnosis biomarkers. Here, we propose 'ExosomePurity', a tumour purity deconvolution model to estimate tumour purity in serum exosomes of cancer patients based on microribonucleic acid (miRNA)-Seq data. We first identify the differently expressed miRNAs as signature to distinguish cancer cell- from healthy cell-derived exosomes. Then, the deconvolution model was developed to estimate the proportions of cancer exosomes and normal exosomes in serum. The purity predicted by the model shows high correlation with actual purity in simulated data and actual data. Moreover, the model is robust under the different levels of noise background. The tumour purity was also used to correct differential expressed gene analysis. ExosomePurity empowers the research community to study non-invasive early diagnosis and to track cancer progression in cancers more efficiently. It is implemented in R and is freely available from GitHub (https://github.com/WangHYLab/ExosomePurity).

The elucidation of gene regulatory networks (GRNs) is one of the central challenges of systems biology, which is crucial for understanding pathogenesis and curing diseases. Various computational methods have been developed for GRN inference, but identifying redundant regulation remains a fundamental problem. Although considering topological properties and edge importance measures simultaneously can identify and reduce redundant regulations, how to address their respective weaknesses whilst leveraging their strengths is a critical problem faced by researchers. Here, we propose a network structure refinement method for GRN (NSRGRN) that effectively combines the topological properties and edge importance measures during GRN inference. NSRGRN has two major parts. The first part constructs a preliminary ranking list of gene regulations to avoid starting the GRN inference from a directed complete graph. The second part develops a novel network structure refinement (NSR) algorithm to refine the network structure from local and global topology perspectives. Specifically, the Conditional Mutual Information with Directionality and network motifs are applied to optimise the local topology, and the lower and upper networks are used to balance the bilateral relationship between the local topology's optimisation and the global topology's maintenance. NSRGRN is compared with six state-of-the-art methods on three datasets (26 networks in total), and it shows the best all-round performance. Furthermore, when acting as a post-processing step, the NSR algorithm can improve the results of other methods in most datasets.

Single-cell ribonucleic acid sequencing (scRNA-seq) enables the quantification of gene expression at the transcriptomic level with single-cell resolution, enhancing our understanding of cellular heterogeneity. However, the excessive missing values present in scRNA-seq data hinder downstream analysis. While numerous imputation methods have been proposed to recover scRNA-seq data, high imputation performance often comes with low or no interpretability. Here, we present IGSimpute, an accurate and interpretable imputation method for recovering missing values in scRNA-seq data with an interpretable instance-wise gene selection layer (GSL). IGSimpute outperforms 12 other state-of-the-art imputation methods on 13 out of 17 datasets from different scRNA-seq technologies with the lowest mean squared error as the chosen benchmark metric. We demonstrate that IGSimpute can give unbiased estimates of the missing values compared to other methods, regardless of whether the average gene expression values are small or large. Clustering results of imputed profiles show that IGSimpute offers statistically significant improvement over other imputation methods. By taking the heart-and-aorta and the limb muscle tissues as examples, we show that IGSimpute can also denoise gene expression profiles by removing outlier entries with unexpectedly high expression values via the instance-wise GSL. We also show that genes selected by the instance-wise GSL could indicate the age of B cells from bladder fat tissue of the Tabula Muris Senis atlas. IGSimpute can impute one million cells using 64 min, and thus applicable to large datasets.

Accurate prediction of promoter regions driving miRNA gene expression has become a major challenge due to the lack of annotation information for pri-miRNA transcripts. This defect hinders our understanding of miRNA-mediated regulatory networks. Some algorithms have been designed during the past decade to detect miRNA promoters. However, these methods rely on biosignal data such as CpG islands and still need to be improved. Here, we propose miProBERT, a BERT-based model for predicting promoters directly from gene sequences without using any structural or biological signals. According to our information, it is the first time a BERT-based model has been employed to identify miRNA promoters. We use the pre-trained model DNABERT, fine-tune the pre-trained model on the gene promoter dataset so that the model includes information about the richer biological properties of promoter sequences in its representation, and then systematically scan the upstream regions of each intergenic miRNA using the fine-tuned model. About, 665 miRNA promoters are found. The innovative use of a random substitution strategy to construct a negative dataset improves the discriminative ability of the model and further reduces the false positive rate (FPR) to as low as 0.0421. On independent datasets, miProBERT outperformed other gene promoter prediction methods. With comparison on 33 experimentally validated miRNA promoter datasets, miProBERT significantly outperformed previously developed miRNA promoter prediction programs with 78.13% precision and 75.76% recall. We further verify the predicted promoter regions by analyzing conservation, CpG content and histone marks. The effectiveness and robustness of miProBERT are highlighted.

Microbiome-based diagnosis of cancer is an increasingly important supplement for the genomics approach in cancer diagnosis, yet current models for microbiome-based diagnosis of cancer face difficulties in generality: not only diagnosis models could not be adapted from one cancer to another, but models built based on microbes from tissues could not be adapted for diagnosis based on microbes from blood. Therefore, a microbiome-based model suitable for a broad spectrum of cancer types is urgently needed. Here we have introduced DeepMicroCancer, a diagnosis model using artificial intelligence techniques for a broad spectrum of cancer types. Built based on the random forest models it has enabled superior performances on more than twenty types of cancers' tissue samples. And by using the transfer learning techniques, improved accuracies could be obtained, especially for cancer types with only a few samples, which could satisfy the requirement in clinical scenarios. Moreover, transfer learning techniques have enabled high diagnosis accuracy that could also be achieved for blood samples. These results indicated that certain sets of microbes could, if excavated using advanced artificial techniques, reveal the intricate differences among cancers and healthy individuals. Collectively, DeepMicroCancer has provided a new venue for accurate diagnosis of cancer based on tissue and blood materials, which could potentially be used in clinics.

Identifying cancer driver genes plays a curial role in the development of precision oncology and cancer therapeutics. Although a plethora of methods have been developed to tackle this problem, the complex cancer mechanisms and intricate interactions between genes still make the identification of cancer driver genes challenging. In this work, we propose a novel machine learning method of heterophilic graph diffusion convolutional networks (called HGDCs) to boost cancer-driver gene identification. Specifically, HGDC first introduces graph diffusion to generate an auxiliary network for capturing the structurally similar nodes in a biomolecular network. Then, HGDC designs an improved message aggregation and propagation scheme to adapt to the heterophilic setting of biomolecular networks, alleviating the problem of driver gene features being smoothed by its neighboring dissimilar genes. Finally, HGDC uses a layer-wise attention classifier to predict the probability of one gene being a cancer driver gene. In the comparison experiments with other existing state-of-the-art methods, our HGDC achieves outstanding performance in identifying cancer driver genes. The experimental results demonstrate that HGDC not only effectively identifies well-known driver genes on different networks but also novel candidate cancer genes. Moreover, HGDC can effectively prioritize cancer driver genes for individual patients. Particularly, HGDC can identify patient-specific additional driver genes, which work together with the well-known driver genes to cooperatively promote tumorigenesis.

Identification of potent peptides through model prediction can reduce benchwork in wet experiments. However, the conventional process of model buildings can be complex and time consuming due to challenges such as peptide representation, feature selection, model selection and hyperparameter tuning. Recently, advanced pretrained deep learning-based language models (LMs) have been released for protein sequence embedding and applied to structure and function prediction. Based on these developments, we have developed UniDL4BioPep, a universal deep-learning model architecture for transfer learning in bioactive peptide binary classification modeling. It can directly assist users in training a high-performance deep-learning model with a fixed architecture and achieve cutting-edge performance to meet the demands in efficiently novel bioactive peptide discovery. To the best of our best knowledge, this is the first time that a pretrained biological language model is utilized for peptide embeddings and successfully predicts peptide bioactivities through large-scale evaluations of those peptide embeddings. The model was also validated through uniform manifold approximation and projection analysis. By combining the LM with a convolutional neural network, UniDL4BioPep achieved greater performances than the respective state-of-the-art models for 15 out of 20 different bioactivity dataset prediction tasks. The accuracy, Mathews correlation coefficient and area under the curve were 0.7-7, 1.23-26.7 and 0.3-25.6% higher, respectively. A user-friendly web server of UniDL4BioPep for the tested bioactivities is established and freely accessible at https://nepc2pvmzy.us-east-1.awsapprunner.com. The source codes, datasets and templates of UniDL4BioPep for other bioactivity fitting and prediction tasks are available at https://github.com/dzjxzyd/UniDL4BioPep.

Prediction of therapy response has been a major challenge in cancer precision medicine due to the extensive tumor heterogeneity. Recently, several deep learning methods have been developed to predict drug response by utilizing various omics data. Most of them train models by using the drug-response screening data generated from cell lines and then use these models to predict response in cancer patient data. In this study, we focus on and evaluate deep learning methods using transcriptome data for the long-standing question of personalized drug-response prediction. We developed an embedding-based approach for drug-response prediction and benchmarked similar methods for their performance. For all methods, we used pretreatment transcriptome data to train models and then conducted a comprehensive evaluation and comparison of the models using cross-panels, cross-datasets and target genes. We further validated the methods using three independent datasets assessing multiple compounds for their predictive capability of drug response, survival outcome and cell line status. As a result, the methods building on gene embeddings had an overall competitive performance with reduced overfitting when we applied evaluation parameters for model fitting as well as the correlation with clinical outcomes in the validation data. We further developed an ensemble model to combine the results from the three most competitive methods for an overall prediction. Finally, we developed DrVAEN (https://bioinfo.uth.edu/drvaen), a user-friendly and easy-accessible web-server that hosts all these methods for drug-response prediction and model comparison for broad use in cancer research, method evaluation and drug development.

While research into drug-target interaction (DTI) prediction is fairly mature, generalizability and interpretability are not always addressed in the existing works in this field. In this paper, we propose a deep learning (DL)-based framework, called BindingSite-AugmentedDTA, which improves drug-target affinity (DTA) predictions by reducing the search space of potential-binding sites of the protein, thus making the binding affinity prediction more efficient and accurate. Our BindingSite-AugmentedDTA is highly generalizable as it can be integrated with any DL-based regression model, while it significantly improves their prediction performance. Also, unlike many existing models, our model is highly interpretable due to its architecture and self-attention mechanism, which can provide a deeper understanding of its underlying prediction mechanism by mapping attention weights back to protein-binding sites. The computational results confirm that our framework can enhance the prediction performance of seven state-of-the-art DTA prediction algorithms in terms of four widely used evaluation metrics, including concordance index, mean squared error, modified squared correlation coefficient ($r^2_m$) and the area under the precision curve. We also contribute to three benchmark drug-traget interaction datasets by including additional information on 3D structure of all proteins contained in those datasets, which include the two most commonly used datasets, namely Kiba and Davis, as well as the data from IDG-DREAM drug-kinase binding prediction challenge. Furthermore, we experimentally validate the practical potential of our proposed framework through in-lab experiments. The relatively high agreement between computationally predicted and experimentally observed binding interactions supports the potential of our framework as the next-generation pipeline for prediction models in drug repurposing.

Molecular clustering analysis has been developed to facilitate visual inspection in the process of structure-based virtual screening. However, traditional methods based on molecular fingerprints or molecular descriptors limit the accuracy of selecting active hit compounds, which may be attributed to the lack of representations of receptor structural and protein-ligand interaction during the clustering. Here, a novel deep clustering framework named ClusterX is proposed to learn molecular representations of protein-ligand complexes and cluster the ligands. In ClusterX, the graph was used to represent the protein-ligand complex, and the joint optimisation can be used efficiently for learning the cluster-friendly features. Experiments on the KLIFs database show that the model can distinguish well between the binding modes of different kinase inhibitors. To validate the effectiveness of the model, the clustering results on the virtual screening dataset further demonstrated that ClusterX achieved better or more competitive performance against traditional methods, such as SIFt and extended connectivity fingerprints. This framework may provide a unique tool for clustering analysis and prove to assist computational medicinal chemists in visual decision-making.

With the development of chromosome conformation capture technique, the study of spatial conformation of a genome based on Hi-C technique has made a quantum leap. Previous studies reveal that genomes are folded into hierarchy of three-dimensional (3D) structures associated with topologically associating domains (TADs), and detecting TAD boundaries is of great significance in the chromosome-level analysis of 3D genome architecture. In this paper, we propose a novel TAD identification method, LPAD, which first extracts node correlations from global interactions of chromosomes based on the random walk with restart and then builds an undirected graph from Hi-C contact matrix. Next, LPAD designs a label propagation-based approach to discover communities and generates TADs. Experimental results verify the effectiveness and quality of TAD detections compared with existing methods. Furthermore, experimental evaluation of chromatin immunoprecipitation sequencing data shows that LPAD performs high enrichment of histone modifications remarkably nearby the TAD boundaries, and these results demonstrate LPAD's advantages on TAD identification accuracy.

Chloroplast is a crucial site for photosynthesis in plants. Determining the location and distribution of proteins in subchloroplasts is significant for studying the energy conversion of chloroplasts and regulating the utilization of light energy in crop production. However, the prediction accuracy of the currently developed protein subcellular site predictors is still limited due to the complex protein sequence features and the scarcity of labeled samples. We propose DaDL-SChlo, a multi-location protein subchloroplast localization predictor, which addresses the above problems by fusing pre-trained protein language model deep learning features with traditional handcrafted features and using generative adversarial networks for data augmentation. The experimental results of cross-validation and independent testing show that DaDL-SChlo has greatly improved the prediction performance of protein subchloroplast compared with the state-of-the-art predictors. Specifically, the overall actual accuracy outperforms the state-of-the-art predictors by 10.7% on 10-fold cross-validation and 12.6% on independent testing. DaDL-SChlo is a promising and efficient predictor for protein subchloroplast localization. The datasets and codes of DaDL-SChlo are available at https://github.com/xwanggroup/DaDL-SChlo.

Identification of active candidate compounds for target proteins, also called drug-protein interaction (DPI) prediction, is an essential but time-consuming and expensive step, which leads to fostering the development of drug discovery. In recent years, deep network-based learning methods were frequently proposed in DPIs due to their powerful capability of feature representation. However, the performance of existing DPI methods is still limited by insufficiently labeled pharmacological data and neglected intermolecular information. Therefore, overcoming these difficulties to perfect the performance of DPIs is an urgent challenge for researchers. In this article, we designed an innovative 'multi-modality attributes' learning-based framework for DPIs with molecular transformer and graph convolutional networks, termed, multi-modality attributes (MMA)-DPI. Specifically, intermolecular sub-structural information and chemical semantic representations were extracted through an augmented transformer module from biomedical data. A tri-layer graph convolutional neural network module was applied to associate the neighbor topology information and learn the condensed dimensional features by aggregating a heterogeneous network that contains multiple biological representations of drugs, proteins, diseases and side effects. Then, the learned representations were taken as the input of a fully connected neural network module to further integrate them in molecular and topological space. Finally, the attribute representations were fused with adaptive learning weights to calculate the interaction score for the DPIs tasks. MMA-DPI was evaluated in different experimental conditions and the results demonstrate that the proposed method achieved higher performance than existing state-of-the-art frameworks.

Human leukocyte antigen class I (HLA-I) molecules bind intracellular peptides produced by protein hydrolysis and present them to the T cells for immune recognition and response. Prediction of peptides that bind HLA-I molecules is very important in immunotherapy. A growing number of computational predictors have been developed in recent years. We survey a comprehensive collection of 27 tools focusing on their input and output data characteristics, key aspects of the underlying predictive models and their availability. Moreover, we evaluate predictive performance for eight representative predictors. We consider a wide spectrum of relevant aspects including allele-specific analysis, influence of negative to positive data ratios and runtime. We also curate high-quality benchmark datasets based on analysis of the consistency of the data labels. Results reveal that each considered method provides accurate results, which can be explained by our analysis that finds that their predictive models capture meaningful binding motifs. Although some methods are overall more accurate than others, we find that none of them is universally superior. We provide a comprehensive comparison of the convenience as well as the accuracy of the methods under specific prediction scenarios, such as for specific alleles, metrics of predictive performance and constraints on runtime. Our systematic and broad analysis provides informative clues to the users to identify the most suitable tools for a given prediction scenario and for the developers to design future methods.

RNA alternative splicing, a post-transcriptional stage in eukaryotes, is crucial in cellular homeostasis and disease processes. Due to the rapid development of the next-generation sequencing (NGS) technology and the flood of NGS data, the detection of differential splicing from RNA-seq data has become mainstream. A range of bioinformatic tools has been developed. However, until now, an independent and comprehensive comparison of available algorithms/tools at the event level is still lacking. Here, 21 different tools are subjected to systematic evaluation, based on simulated RNA-seq data where exact differential splicing events are introduced. We observe immense discrepancies among these tools. SUPPA, DARTS, rMATS and LeafCutter outperforme other event-based tools. We also examine the abilities of the tools to identify novel splicing events, which shows that most event-based tools are unsuitable for discovering novel splice sites. To improve the overall performance, we present two methodological approaches i.e. low-expression transcript filtering and tool-pair combination. Finally, a new protocol of selecting tools to perform differential splicing analysis for different analytical tasks (e.g. precision and recall rate) is proposed. Under this protocol, we analyze the distinct splicing landscape in the DUX4/IGH subgroup of B-cell acute lymphoblastic leukemia and uncover the differential splicing of TCF12. All codes needed to reproduce the results are available at https://github.com/mhjiang97/Benchmarking_DS.

Small open reading frames (smORFs) encoding proteins less than 100 amino acids (aa) are known to be important regulators of key cellular processes. However, their computational identification remains a challenge. Based on a comprehensive analysis of known prokaryotic small ORFs, we have developed the ProsmORF-pred resource which uses a machine learning (ML)-based method for prediction of smORFs in the prokaryotic genome sequences. ProsmORF-pred consists of two ML models, one for initiation site recognition in nucleic acid sequences upstream of putative start codons and the other uses translated amino acid sequences to decipher functional protein like sequences. The nucleotide sequence-based initiation site recognition model has been trained using longer ORFs (>100 aa) in the same genome while the ML model for identification of protein like sequences has been trained using annotated smORFs from Escherichia coli. Comprehensive benchmarking of ProsmORF-pred reveals that its performance is comparable to other state-of-the-art approaches on the annotated smORF set derived from 32 prokaryotic genomes. Its performance is distinctly superior to other tools like PRODIGAL and RANSEPS for prediction of newly identified smORFs which have a length range of 10-30 aa, where prediction of smORFs has been a major challenge. Apart from identification of smORFs in genomic sequences, ProsmORF-pred can also aid in functional annotation of the predicted smORFs based on sequence similarity and genomic neighbourhood similarity searches in ProsmORFDB, a well-curated database of known smORFs. ProsmORF-pred along with its backend database ProsmORFDB is available as a user-friendly web server (http://www.nii.ac.in/prosmorfpred.html).

Undoubtedly, single-cell RNA sequencing (scRNA-seq) has changed the research landscape by providing insights into heterogeneous, complex and rare cell populations. Given that more such data sets will become available in the near future, their accurate assessment with compatible and robust models for cell type annotation is a prerequisite. Considering this, herein, we developed scAnno (scRNA-seq data annotation), an automated annotation tool for scRNA-seq data sets primarily based on the single-cell cluster levels, using a joint deconvolution strategy and logistic regression. We explicitly constructed a reference profile for human (30 cell types and 50 human tissues) and a reference profile for mouse (26 cell types and 50 mouse tissues) to support this novel methodology (scAnno). scAnno offers a possibility to obtain genes with high expression and specificity in a given cell type as cell type-specific genes (marker genes) by combining co-expression genes with seed genes as a core. Of importance, scAnno can accurately identify cell type-specific genes based on cell type reference expression profiles without any prior information. Particularly, in the peripheral blood mononuclear cell data set, the marker genes identified by scAnno showed cell type-specific expression, and the majority of marker genes matched exactly with those included in the CellMarker database. Besides validating the flexibility and interpretability of scAnno in identifying marker genes, we also proved its superiority in cell type annotation over other cell type annotation tools (SingleR, scPred, CHETAH and scmap-cluster) through internal validation of data sets (average annotation accuracy: 99.05%) and cross-platform data sets (average annotation accuracy: 95.56%). Taken together, we established the first novel methodology that utilizes a deconvolution strategy for automated cell typing and is capable of being a significant application in broader scRNA-seq analysis. scAnno is available at https://github.com/liuhong-jia/scAnno.

Determining the interacting proteins in multiprotein complexes can be technically challenging. An emerging biochemical approach to this end is based on the 'thermal proximity co-aggregation' (TPCA) phenomenon. Accordingly, when two or more proteins interact to form a complex, they tend to co-aggregate when subjected to heat-induced denaturation and thus exhibit similar melting curves. Here, we explore the potential of leveraging TPCA for determining protein interactions. We demonstrate that dissimilarity measure-based information retrieval applied to melting curves tends to rank a protein-of-interest's interactors higher than its non-interactors, as shown in the context of pull-down assay results. Consequently, such rankings can reduce the number of confirmatory biochemical experiments needed to find bona fide protein-protein interactions. In general, rankings based on dissimilarity measures generated through metric learning further reduce the required number of experiments compared to those based on standard dissimilarity measures such as Euclidean distance. When a protein mixture's melting curves are obtained in two conditions, we propose a scoring function that uses melting curve data to inform how likely a protein pair is to interact in one condition but not another. We show that ranking protein pairs by their scores is an effective approach for determining condition-specific protein-protein interactions. By contrast, clustering melting curve data generally does not inform about the interacting proteins in multiprotein complexes. In conclusion, we report improved methods for dissimilarity measure-based computation of melting curves data that can greatly enhance the determination of interacting proteins in multiprotein complexes.

Cyclic AMP receptor proteins (CRPs) are important transcription regulators in many species. The prediction of CRP-binding sites was mainly based on position-weighted matrixes (PWMs). Traditional prediction methods only considered known binding motifs, and their ability to discover inflexible binding patterns was limited. Thus, a novel CRP-binding site prediction model called CRPBSFinder was developed in this research, which combined the hidden Markov model, knowledge-based PWMs and structure-based binding affinity matrixes. We trained this model using validated CRP-binding data from Escherichia coli and evaluated it with computational and experimental methods. The result shows that the model not only can provide higher prediction performance than a classic method but also quantitatively indicates the binding affinity of transcription factor binding sites by prediction scores. The prediction result included not only the most knowns regulated genes but also 1089 novel CRP-regulated genes. The major regulatory roles of CRPs were divided into four classes: carbohydrate metabolism, organic acid metabolism, nitrogen compound metabolism and cellular transport. Several novel functions were also discovered, including heterocycle metabolic and response to stimulus. Based on the functional similarity of homologous CRPs, we applied the model to 35 other species. The prediction tool and the prediction results are online and are available at: https://awi.cuhk.edu.cn/∼CRPBSFinder.

Studies have found that human microbiome is associated with and predictive of human health and diseases. Many statistical methods developed for microbiome data focus on different distance metrics that can capture various information in microbiomes. Prediction models were also developed for microbiome data, including deep learning methods with convolutional neural networks that consider both taxa abundance profiles and taxonomic relationships among microbial taxa from a phylogenetic tree. Studies have also suggested that a health outcome could associate with multiple forms of microbiome profiles. In addition to the abundance of some taxa that are associated with a health outcome, the presence/absence of some taxa is also associated with and predictive of the same health outcome. Moreover, associated taxa may be close to each other on a phylogenetic tree or spread apart on a phylogenetic tree. No prediction models currently exist that use multiple forms of microbiome-outcome associations. To address this, we propose a multi-kernel machine regression (MKMR) method that is able to capture various types of microbiome signals when doing predictions. MKMR utilizes multiple forms of microbiome signals through multiple kernels being transformed from multiple distance metrics for microbiomes and learn an optimal conic combination of these kernels, with kernel weights helping us understand contributions of individual microbiome signal types. Simulation studies suggest a much-improved prediction performance over competing methods with mixture of microbiome signals. Real data applicants to predict multiple health outcomes using throat and gut microbiome data also suggest a better prediction of MKMR than that of competing methods.

A-to-I editing is the most prevalent RNA editing event, which refers to the change of adenosine (A) bases to inosine (I) bases in double-stranded RNAs. Several studies have revealed that A-to-I editing can regulate cellular processes and is associated with various human diseases. Therefore, accurate identification of A-to-I editing sites is crucial for understanding RNA-level (i.e. transcriptional) modifications and their potential roles in molecular functions. To date, various computational approaches for A-to-I editing site identification have been developed; however, their performance is still unsatisfactory and needs further improvement. In this study, we developed a novel stacked-ensemble learning model, ATTIC (A-To-I ediTing predICtor), to accurately identify A-to-I editing sites across three species, including Homo sapiens, Mus musculus and Drosophila melanogaster. We first comprehensively evaluated 37 RNA sequence-derived features combined with 14 popular machine learning algorithms. Then, we selected the optimal base models to build a series of stacked ensemble models. The final ATTIC framework was developed based on the optimal models improved by the feature selection strategy for specific species. Extensive cross-validation and independent tests illustrate that ATTIC outperforms state-of-the-art tools for predicting A-to-I editing sites. We also developed a web server for ATTIC, which is publicly available at http://web.unimelb-bioinfortools.cloud.edu.au/ATTIC/. We anticipate that ATTIC can be utilized as a useful tool to accelerate the identification of A-to-I RNA editing events and help characterize their roles in post-transcriptional regulation.

Traditional Chinese medicine (TCM) has accumulated thousands years of knowledge in herbal therapy, but the use of herbal formulas is still characterized by reliance on personal experience. Due to the complex mechanism of herbal actions, it is challenging to discover effective herbal formulas for diseases by integrating the traditional experiences and modern pharmacological mechanisms of multi-target interactions. In this study, we propose a herbal formula prediction approach (TCMFP) combined therapy experience of TCM, artificial intelligence and network science algorithms to screen optimal herbal formula for diseases efficiently, which integrates a herb score (Hscore) based on the importance of network targets, a pair score (Pscore) based on empirical learning and herbal formula predictive score (FmapScore) based on intelligent optimization and genetic algorithm. The validity of Hscore, Pscore and FmapScore was verified by functional similarity and network topological evaluation. Moreover, TCMFP was used successfully to generate herbal formulae for three diseases, i.e. the Alzheimer's disease, asthma and atherosclerosis. Functional enrichment and network analysis indicates the efficacy of targets for the predicted optimal herbal formula. The proposed TCMFP may provides a new strategy for the optimization of herbal formula, TCM herbs therapy and drug development.

Cardiac conduction disease is a major cause of morbidity and mortality worldwide. There is considerable clinical significance and an emerging need of early detection of these diseases for preventive treatment success before more severe arrhythmias occur. However, developing such early screening tools is challenging due to the lack of early electrocardiograms (ECGs) before symptoms occur in patients. Mouse models are widely used in cardiac arrhythmia research. The goal of this paper is to develop deep learning models to predict cardiac conduction diseases in mice using their early ECGs. We hypothesize that mutant mice present subtle abnormalities in their early ECGs before severe arrhythmias present. These subtle patterns can be detected by deep learning though they are hard to be identified by human eyes. We propose a deep transfer learning model, DeepMiceTL, which leverages knowledge from human ECGs to learn mouse ECG patterns. We further apply the Bayesian optimization and $k$-fold cross validation methods to tune the hyperparameters of the DeepMiceTL. Our results show that DeepMiceTL achieves a promising performance (F1-score: 83.8%, accuracy: 84.8%) in predicting the occurrence of cardiac conduction diseases using early mouse ECGs. This study is among the first efforts that use state-of-the-art deep transfer learning to identify ECG patterns during the early course of cardiac conduction disease in mice. Our approach not only could help in cardiac conduction disease research in mice, but also suggest a feasibility for early clinical diagnosis of human cardiac conduction diseases and other types of cardiac arrythmias using deep transfer learning in the future.

PURPOSE: Evaluation of genetic mutations in cancers is important because distinct mutational profiles help determine individualized drug therapy. However, molecular analyses are not routinely performed in all cancers because they are expensive, time-consuming and not universally available. Artificial intelligence (AI) has shown the potential to determine a wide range of genetic mutations on histologic image analysis. Here, we assessed the status of mutation prediction AI models on histologic images by a systematic review. METHODS: A literature search using the MEDLINE, Embase and Cochrane databases was conducted in August 2021. The articles were shortlisted by titles and abstracts. After a full-text review, publication trends, study characteristic analysis and comparison of performance metrics were performed. RESULTS: Twenty-four studies were found mostly from developed countries, and their number is increasing. The major targets were gastrointestinal, genitourinary, gynecological, lung and head and neck cancers. Most studies used the Cancer Genome Atlas, with a few using an in-house dataset. The area under the curve of some of the cancer driver gene mutations in particular organs was satisfactory, such as 0.92 of BRAF in thyroid cancers and 0.79 of EGFR in lung cancers, whereas the average of all gene mutations was 0.64, which is still suboptimal. CONCLUSION: AI has the potential to predict gene mutations on histologic images with appropriate caution. Further validation with larger datasets is still required before AI models can be used in clinical practice to predict gene mutations.

Single-cell RNA sequencing (scRNA-seq) has significantly accelerated the experimental characterization of distinct cell lineages and types in complex tissues and organisms. Cell-type annotation is of great importance in most of the scRNA-seq analysis pipelines. However, manual cell-type annotation heavily relies on the quality of scRNA-seq data and marker genes, and therefore can be laborious and time-consuming. Furthermore, the heterogeneity of scRNA-seq datasets poses another challenge for accurate cell-type annotation, such as the batch effect induced by different scRNA-seq protocols and samples. To overcome these limitations, here we propose a novel pipeline, termed TripletCell, for cross-species, cross-protocol and cross-sample cell-type annotation. We developed a cell embedding and dimension-reduction module for the feature extraction (FE) in TripletCell, namely TripletCell-FE, to leverage the deep metric learning-based algorithm for the relationships between the reference gene expression matrix and the query cells. Our experimental studies on 21 datasets (covering nine scRNA-seq protocols, two species and three tissues) demonstrate that TripletCell outperformed state-of-the-art approaches for cell-type annotation. More importantly, regardless of protocols or species, TripletCell can deliver outstanding and robust performance in annotating different types of cells. TripletCell is freely available at https://github.com/liuyan3056/TripletCell. We believe that TripletCell is a reliable computational tool for accurately annotating various cell types using scRNA-seq data and will be instrumental in assisting the generation of novel biological hypotheses in cell biology.

Protein function prediction is an essential task in bioinformatics which benefits disease mechanism elucidation and drug target discovery. Due to the explosive growth of proteins in sequence databases and the diversity of their functions, it remains challenging to fast and accurately predict protein functions from sequences alone. Although many methods have integrated protein structures, biological networks or literature information to improve performance, these extra features are often unavailable for most proteins. Here, we propose SPROF-GO, a Sequence-based alignment-free PROtein Function predictor, which leverages a pretrained language model to efficiently extract informative sequence embeddings and employs self-attention pooling to focus on important residues. The prediction is further advanced by exploiting the homology information and accounting for the overlapping communities of proteins with related functions through the label diffusion algorithm. SPROF-GO was shown to surpass state-of-the-art sequence-based and even network-based approaches by more than 14.5, 27.3 and 10.1% in area under the precision-recall curve on the three sub-ontology test sets, respectively. Our method was also demonstrated to generalize well on non-homologous proteins and unseen species. Finally, visualization based on the attention mechanism indicated that SPROF-GO is able to capture sequence domains useful for function prediction. The datasets, source codes and trained models of SPROF-GO are available at https://github.com/biomed-AI/SPROF-GO. The SPROF-GO web server is freely available at http://bio-web1.nscc-gz.cn/app/sprof-go.

Gene essentiality is defined as the extent to which a gene is required for the survival and reproductive success of a living system. It can vary between genetic backgrounds and environments. Essential protein coding genes have been well studied. However, the essentiality of non-coding regions is rarely reported. Most regions of human genome do not encode proteins. Determining essentialities of non-coding genes is demanded. We developed iEssLnc models, which can assign essentiality scores to lncRNA genes. As far as we know, this is the first direct quantitative estimation to the essentiality of lncRNA genes. By taking the advantage of graph neural network with meta-path-guided random walks on the lncRNA-protein interaction network, iEssLnc models can perform genome-wide screenings for essential lncRNA genes in a quantitative manner. We carried out validations and whole genome screening in the context of human cancer cell-lines and mouse genome. In comparisons to other methods, which are transferred from protein-coding genes, iEssLnc achieved better performances. Enrichment analysis indicated that iEssLnc essentiality scores clustered essential lncRNA genes with high ranks. With the screening results of iEssLnc models, we estimated the number of essential lncRNA genes in human and mouse. We performed functional analysis to find that essential lncRNA genes interact with microRNAs and cytoskeletal proteins significantly, which may be of interest in experimental life sciences. All datasets and codes of iEssLnc models have been deposited in GitHub (https://github.com/yyZhang14/iEssLnc).

Determining drug-drug interactions (DDIs) is an important part of pharmacovigilance and has a vital impact on public health. Compared with drug trials, obtaining DDI information from scientific articles is a faster and lower cost but still a highly credible approach. However, current DDI text extraction methods consider the instances generated from articles to be independent and ignore the potential connections between different instances in the same article or sentence. Effective use of external text data could improve prediction accuracy, but existing methods cannot extract key information from external data accurately and reasonably, resulting in low utilization of external data. In this study, we propose a DDI extraction framework, instance position embedding and key external text for DDI (IK-DDI), which adopts instance position embedding and key external text to extract DDI information. The proposed framework integrates the article-level and sentence-level position information of the instances into the model to strengthen the connections between instances generated from the same article or sentence. Moreover, we introduce a comprehensive similarity-matching method that uses string and word sense similarity to improve the matching accuracy between the target drug and external text. Furthermore, the key sentence search method is used to obtain key information from external data. Therefore, IK-DDI can make full use of the connection between instances and the information contained in external text data to improve the efficiency of DDI extraction. Experimental results show that IK-DDI outperforms existing methods on both macro-averaged and micro-averaged metrics, which suggests our method provides complete framework that can be used to extract relationships between biomedical entities and process external text data.

Identifying protein-protein interaction (PPI) site is an important step in understanding biological activity, apprehending pathological mechanism and designing novel drugs. Developing reliable computational methods for predicting PPI site as screening tools contributes to reduce lots of time and expensive costs for conventional experiments, but how to improve the accuracy is still challenging. We propose a PPI site predictor, called Augmented Graph Attention Network Protein-Protein Interacting Site (AGAT-PPIS), based on AGAT with initial residual and identity mapping, in which eight AGAT layers are connected to mine node embedding representation deeply. AGAT is our augmented version of graph attention network, with added edge features. Besides, extra node features and edge features are introduced to provide more structural information and increase the translation and rotation invariance of the model. On the benchmark test set, AGAT-PPIS significantly surpasses the state-of-the-art method by 8% in Accuracy, 17.1% in Precision, 11.8% in F1-score, 15.1% in Matthews Correlation Coefficient (MCC), 8.1% in Area Under the Receiver Operating Characteristic curve (AUROC), 14.5% in Area Under the Precision-Recall curve (AUPRC), respectively.

MOTIVATION: A large number of studies have shown that circular RNA (circRNA) affects biological processes by competitively binding miRNA, providing a new perspective for the diagnosis, and treatment of human diseases. Therefore, exploring the potential circRNA-miRNA interactions (CMIs) is an important and urgent task at present. Although some computational methods have been tried, their performance is limited by the incompleteness of feature extraction in sparse networks and the low computational efficiency of lengthy data. RESULTS: In this paper, we proposed JSNDCMI, which combines the multi-structure feature extraction framework and Denoising Autoencoder (DAE) to meet the challenge of CMI prediction in sparse networks. In detail, JSNDCMI integrates functional similarity and local topological structure similarity in the CMI network through the multi-structure feature extraction framework, then forces the neural network to learn the robust representation of features through DAE and finally uses the Gradient Boosting Decision Tree classifier to predict the potential CMIs. JSNDCMI produces the best performance in the 5-fold cross-validation of all data sets. In the case study, seven of the top 10 CMIs with the highest score were verified in PubMed. AVAILABILITY: The data and source code can be found at https://github.com/1axin/JSNDCMI.

Riboswitches are conserved structural ribonucleic acid (RNA) sensors that are mainly found to regulate a large number of genes/operons in bacteria. Presently, >50 bacterial riboswitch classes have been discovered, but only the thiamine pyrophosphate riboswitch class is detected in a few eukaryotes like fungi, plants and algae. One of the most important challenges in riboswitch research is to discover existing riboswitch classes in eukaryotes and to understand the evolution of bacterial riboswitches. However, traditional search methods for riboswitch detection have failed to detect eukaryotic riboswitches besides just one class and any distant structural homologs of riboswitches. We developed a novel approach based on inverse RNA folding that attempts to find sequences that match the shape of the target structure with minimal sequence conservation based on key nucleotides that interact directly with the ligand. Then, to support our matched candidates, we expanded the results into a covariance model representing similar sequences preserving the structure. Our method transforms a structure-based search into a sequence-based search that considers the conservation of secondary structure shape and ligand-binding residues. This method enables us to identify a potential structural candidate in fungi that could be the distant homolog of bacterial purine riboswitches. Further, phylogenomic analysis and evolutionary distribution of this structural candidate indicate that the most likely point of origin of this structural candidate in these organisms is associated with the loss of traditional purine riboswitches. The computational approach could be applicable to other domains and problems in RNA research.

Post- and co-transcriptional RNA modifications are found to play various roles in regulating essential biological processes at all stages of RNA life. Precise identification of RNA modification sites is thus crucial for understanding the related molecular functions and specific regulatory circuitry. To date, a number of computational approaches have been developed for in silico identification of RNA modification sites; however, most of them require learning from base-resolution epitranscriptome datasets, which are generally scarce and available only for a limited number of experimental conditions, and predict only a single modification, even though there are multiple inter-related RNA modification types available. In this study, we proposed AdaptRM, a multi-task computational method for synergetic learning of multi-tissue, type and species RNA modifications from both high- and low-resolution epitranscriptome datasets. By taking advantage of adaptive pooling and multi-task learning, the newly proposed AdaptRM approach outperformed the state-of-the-art computational models (WeakRM and TS-m6A-DL) and two other deep-learning architectures based on Transformer and ConvMixer in three different case studies for both high-resolution and low-resolution prediction tasks, demonstrating its effectiveness and generalization ability. In addition, by interpreting the learned models, we unveiled for the first time the potential association between different tissues in terms of epitranscriptome sequence patterns. AdaptRM is available as a user-friendly web server from http://www.rnamd.org/AdaptRM together with all the codes and data used in this project.

Single-cell RNA-seq analysis has become a powerful tool to analyse the transcriptomes of individual cells. In turn, it has fostered the possibility of screening thousands of single cells in parallel. Thus, contrary to the traditional bulk measurements that only paint a macroscopic picture, gene measurements at the cell level aid researchers in studying different tissues and organs at various stages. However, accurate clustering methods for such high-dimensional data remain exiguous and a persistent challenge in this domain. Of late, several methods and techniques have been promulgated to address this issue. In this article, we propose a novel framework for clustering large-scale single-cell data and subsequently identifying the rare-cell sub-populations. To handle such sparse, high-dimensional data, we leverage PaCMAP (Pairwise Controlled Manifold Approximation), a feature extraction algorithm that preserves both the local and the global structures of the data and Gaussian Mixture Model to cluster single-cell data. Subsequently, we exploit Edited Nearest Neighbours sampling and Isolation Forest/One-class Support Vector Machine to identify rare-cell sub-populations. The performance of the proposed method is validated using the publicly available datasets with varying degrees of cell types and rare-cell sub-populations. On several benchmark datasets, the proposed method outperforms the existing state-of-the-art methods. The proposed method successfully identifies cell types that constitute populations ranging from 0.1 to 8% with F1-scores of 0.91 0.09. The source code is available at https://github.com/scrab017/RarPG.

MOTIVATION: Single-cell RNA sequencing (scRNA-seq) technology attracts extensive attention in the biomedical field. It can be used to measure gene expression and analyze the transcriptome at the single-cell level, enabling the identification of cell types based on unsupervised clustering. Data imputation and dimension reduction are conducted before clustering because scRNA-seq has a high 'dropout' rate, noise and linear inseparability. However, independence of dimension reduction, imputation and clustering cannot fully characterize the pattern of the scRNA-seq data, resulting in poor clustering performance. Herein, we propose a novel and accurate algorithm, SSNMDI, that utilizes a joint learning approach to simultaneously perform imputation, dimensionality reduction and cell clustering in a non-negative matrix factorization (NMF) framework. In addition, we integrate the cell annotation as prior information, then transform the joint learning into a semi-supervised NMF model. Through experiments on 14 datasets, we demonstrate that SSNMDI has a faster convergence speed, better dimensionality reduction performance and a more accurate cell clustering performance than previous methods, providing an accurate and robust strategy for analyzing scRNA-seq data. Biological analysis are also conducted to validate the biological significance of our method, including pseudotime analysis, gene ontology and survival analysis. We believe that we are among the first to introduce imputation, partial label information, dimension reduction and clustering to the single-cell field. AVAILABILITY AND IMPLEMENTATION: The source code for SSNMDI is available at https://github.com/yushanqiu/SSNMDI.