The microRNAs are non-coding RNAs which post-transcriptionally regulate the expression of many eukaryotic genes, and whose dysregulation is a driver of human disease. Here we report the discovery of a very slow (0.1 s-1) conformational rearrangement at the Dicer cleavage site of pre-miR-21, which regulates the relative concentration of readily- and inefficiently-processed RNA structural states. We show that this dynamic switch is affected by single nucleotide mutations and can be biased by small molecule and peptide ligands, which can direct the microRNA to occupy the inefficiently processed state and reduce processing efficiency. This result reveals a new mechanism of RNA regulation and suggests a chemical approach to suppressing or activating pathogenic microRNAs by selective stabilization of their unprocessed or processed states.

Nucleosome reconstitution plays an important role in many cellular functions. As an initial step, H2A-H2B dimer displacement, which is accompanied by disruption of many of the interactions within the nucleosome, should occur. To understand how H2A-H2B dimer displacement occurs, an adaptively biased molecular dynamics (ABMD) simulation was carried out to generate a variety of displacements of the H2A-H2B dimer from the fully wrapped to partially unwrapped nucleosome structures. With regards to these structures, the free energy landscape of the dimer displacement was investigated using umbrella sampling simulations. We found that the main contributors to the free energy were the docking domain of H2A and the C-terminal of H4. There were various paths for the dimer displacement which were dependent on the extent of nucleosomal DNA wrapping, suggesting that modulation of the intra-nucleosomal interaction by external factors such as histone chaperones could control the path for the H2A-H2B dimer displacement. Key residues which contributed to the free energy have also been reported to be involved in the mutations and posttranslational modifications (PTMs) which are important for assembling and/or reassembling the nucleosome at the molecular level and are found in cancer cells at the phenotypic level. Our results give insight into how the H2A-H2B dimer displacement proceeds along various paths according to different interactions within the nucleosome.

Disordered scaffold proteins provide multivalent landing pads that, via a series of embedded Short Linear Motifs (SLiMs), bring together the components of a complex to orchestrate precise spatial and temporal regulation of cellular processes. One such protein is AKAP5 (previously AKAP79), which contains SLiMs that anchor PKA and Calcineurin, and recruit substrate (the TRPV1 receptor). Calcineurin is anchored to AKAP5 by a well-characterised PxIxIT SLiM. Here we show, using a combination of biochemical and biophysical approaches, that the Calcineurin PxIxIT-binding groove also recognises several hitherto unknown lower-affinity SLiMs in addition to the PxIxIT motif. We demonstrate that the assembly is in reality a complex system with conserved SLiMs spanning a wide affinity range. The capture is analogous to that seen for many DNA-binding proteins that have a weak non-specific affinity for DNA outside the canonical binding site, but different in that it involves (i) two proteins, and (ii) hydrophobic rather than electrostatic interactions. It is also compatible with the requirement for both stable anchoring of the enzyme and responsive downstream signalling. We conclude that the AKAP5 C-terminus is enriched in lower-affinity/mini-SLiMs that, together with the canonical SLiM, maintain a structurally disordered but tightly regulated signalosome.

Staphylococcus epidermidis is a commensal bacterium on human skin that is also the leading cause of medical device-related infections. The accumulation-associated protein (Aap) from S. epidermidis is a critical factor for infection via its ability to mediate biofilm formation. The B-repeat superdomain of Aap is composed of 5 to 17 Zn2+-binding B-repeats, which undergo rapid, reversible assembly to form dimer and tetramer species. The tetramer can then undergo a conformational change and nucleate highly stable functional amyloid fibrils. In this study, multiple techniques including analytical ultracentrifugation (AUC) and small-angle X-ray scattering (SAXS) are used to probe a panel of B-repeat mutant constructs that assemble to distinct oligomeric states to define the structural characteristics of B-repeat dimer and tetramer species. The B-repeat region from Aap forms an extremely elongated conformation that presents several challenges for standard SAXS analyses. Specialized approaches, such as cross-sectional analyses, allowed for in-depth interpretation of data, while explicit-solvent calculations via WAXSiS allowed for accurate evaluation of atomistic models. The resulting models suggest mechanisms by which Aap functional amyloid fibrils form, illuminating an important contributing factor to recurrent staphylococcal infections.

As a member of PDGF/VEGF (Platelet-derived growth factor/ Vascular endothelial growth factor) growth factors, PDGF-D regulates blood vessel development, wound healing, innate immunity, and organogenesis. Unlike PDGF-A and PDGF-B, PDGF-D has an additional CUB (Complement C1r/C1s, Uegf, Bmp1) domain at the N-terminus of its growth factor domain, and thus it is secreted in a latent, inactive complex, which needs to be proteolytically activated for its biological activities. However, how the CUB domain contributes to the latency and activation of the growth factor remains elusive. In this study, we modeled the dimeric structure of PDGF-D pro-complex and studied the inhibitory functions of PDGF-D prodomain on PDGF-B and PDGF-D signaling. In our model, the growth factor domain of PDGF-D forms a VEGF-D-like dimer through their β1 and β3 interactions. The hinge and CUB domains of PDGF-D bind at the opposite sides of the growth factor domain and exclude the PDGFR-β (PDGF Receptor β) D2 and D3 domains from recognizing the growth factor. In addition, we verified that PDGF-D prodomain could inhibit both PDGF-B and PDGF-D mediated PDGFR-β transphosphorylation in a dose-dependent manner. However, PDGF-D prodomain could only inhibit the proliferation of NIH 3T3 cells stimulated by PDGF-D but not by PDGF-B, indicating its differential inhibitory activities toward PDGF-B and PDGF-D signaling.

Solution and solid-state NMR spectroscopy are highly complementary techniques for studying structure and dynamics in very high molecular weight systems. Here we have analysed the dynamics of HIV-1 capsid (CA) assemblies in presence of the cofactors IP6 and ATPγS and the host-factor CPSF6 using a combination of solution state and cross polarisation magic angle spinning (CP-MAS) solid-state NMR. In particular, dynamical effects on ns to µs and µs to ms timescales are observed revealing diverse motions in assembled CA. Using CP-MAS NMR, we exploited the sensitivity of the amide/Cα-Cβ backbone chemical shifts in DARR and NCA spectra to observe the plasticity of the HIV-1 CA tubular assemblies and also map the binding of cofactors and the dynamics of cofactor-CA complexes. In solution, we measured how the addition of host- and co-factors to CA -hexamers perturbed the chemical shifts and relaxation properties of CA-Ile and -Met methyl groups using transverse-relaxation-optimized NMR spectroscopy to exploit the sensitivity of methyl groups as probes in high-molecular weight proteins. These data show how dynamics of the CA protein assembly over a range of spatial and temporal scales play a critical role in CA function. Moreover, we show that binding of IP6, ATPγS and CPSF6 results in local chemical shift as well as dynamic changes for a significant, contiguous portion of CA, highlighting how allosteric pathways communicate ligand interactions between adjacent CA protomers.

Viral infection in cells triggers a cascade of molecular defense mechanisms to maintain host-cell homoeostasis. One of these mechanisms is ADP-ribosylation, a fundamental post-translational modification (PTM) characterized by the addition of ADP-ribose (ADPr) on substrates. Poly(ADP-ribose) polymerases (PARPs) are implicated in this process and they perform ADP-ribosylation on host and pathogen proteins. Some viral families contain structural motifs that can reverse this PTM. These motifs known as macro domains (MDs) are evolutionarily conserved protein domains found in all kingdoms of life. They are divided in different classes with the viral belonging to Macro-D-type class because of their properties to recognize and revert the ADP-ribosylation. Viral MDs are potential pharmaceutical targets, capable to counteract host immune response. Sequence and structural homology between viral and human MDs are an impediment for the development of new active compounds against their function. Remdesivir, is a drug administrated in viral infections inhibiting viral replication through RNA-dependent RNA polymerase (RdRp). Herein, GS-441524, the active metabolite of the remdesivir, is tested as a hydrolase inhibitor for several viral MDs and for its binding to human homologs found in PARPs. This study presents biochemical and biophysical studies, which indicate that GS-441524 selectively modifies SARS-CoV-2 MD de-MARylation activity, while it does not interact with hPARP14 MD2 and hPARP15 MD2. The structural investigation of MD•GS-441524 complexes, using solution NMR and X-ray crystallography, discloses the impact of certain amino acids in ADPr binding cavity suggesting that F360 and its adjacent residues tune the selective binding of the inhibitor to SARS-CoV-2 MD.

Cell-free expression enables direct cotranslational insertion of G protein coupled receptors (GPCRs) and other membrane proteins into the defined membrane environments of nanodiscs. This technique avoids GPCR contacts with detergents and allows rapid identification of lipid effects on GPCR function as well as fast screening of receptor derivatives. Critical steps of conventional GPCR preparation from cellular membranes followed by detergent-based reconstitution into nanodisc membranes are thus eliminated. We report the efficient cotranslational insertion of full-length human β1-adrenergic receptor and of a truncated derivative into preformed nanodisc membranes. Their biochemical characterization revealed significant differences in lipid requirements, dimer formation and ligand binding activity. The truncated receptor showed a higher affinity to most tested ligands, in particular in presence of choline-containing lipids. However, introducing the naturally occurring G389R polymorphism in the full-length receptor resulted into an increased affinity to the antagonists alprenolol and carvedilol. Receptor quality was generally improved by coexpression with the agonist isoproterenol and the percentage of the ligand binding active fraction was twofold increased. Specific coupling of full-length and truncated human receptors in nanodisc membranes to Mini-Gαs protein as well as to purified Gs heterotrimer could be demonstrated and homogeneity of purified GPCR/Gs protein complexes in nanodiscs was demonstrated by negative stain single particle analysis.

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease primarily impacting motor neurons. Mutations in superoxide dismutase 1 (SOD1) are the second most common cause of familial ALS. Several of these mutations lead to misfolding or toxic gain of function in the SOD1 protein. Recently, we reported that misfolded SOD1 interacts with TNF receptor-associated factor 6 (TRAF6) in the SOD1G93A rat model of ALS. Further, we showed in cultured cells that several mutant SOD1 proteins, but not wildtype SOD1 protein, interact with TRAF6 via the MATH domain. Here, we sought to uncover the structural details of this interaction through molecular dynamics (MD) simulations of a dimeric model system, coarse grained using the AWSEM force field. We used direct MD simulations to identify buried residues, and predict binding poses by clustering frames from the trajectories. Metadynamics simulations were also used to deduce preferred binding regions on the protein surfaces from the potential of the mean force in orientation space. Well-folded SOD1 was found to bind TRAF6 via co-option of its native homodimer interface. However, if loops IV and VII of SOD1 were disordered, as typically occurs in the absence of stabilizing Zn2+ ion binding, these disordered loops now participated in novel interactions with TRAF6. On TRAF6, multiple interaction hot-spots were distributed around the equatorial region of the MATH domain beta barrel. Expression of TRAF6 variants with mutations in this region in cultured cells demonstrated that TRAF6T475 facilitates interaction with different SOD1 mutants. These findings contribute to our understanding of the disease mechanism and uncover potential targets for the development of therapeutics.

Telomeres are nucleoprotein complexes that protect the ends of chromosomes and are essential for chromosome stability in Eukaryotes. In cells, individual telomeres form distinct globules of finite size that appear to be smaller than expected for bare DNA. Moreover, telomeres can cluster together, form telomere-induced-foci or co-localize with promyelocytic leukemia (PML) nuclear bodies. The physical basis for collapse of individual telomeres and coalescence of multiple ones remains unclear, as does the relationship between these two phenomena. By combining single-molecule force spectroscopy measurements, optical microscopy, turbidity assays, and simulations, we show that the telomere scaffolding protein TRF2 can condense individual DNA chains and drives coalescence of multiple DNA molecules, leading to phase separation and the formation of liquid-like droplets. Addition of the TRF2 binding protein hRap1 modulates phase boundaries and tunes the specificity of solution demixing while simultaneously altering the degree of DNA compaction. Our results suggest that the condensation of single telomeres and formation of biomolecular condensates containing multiple telomeres are two different outcomes driven by the same set of molecular interactions. Moreover, binding partners, such as other telomere components, can alter those interactions to promote single-chain DNA compaction over multiple-chain phase separation.

The envelope glycoprotein gp41 of the HIV-1 virus mediates its entry into the host cell. During this process, gp41 undergoes large conformational changes and the energy released in the remodeling events is utilized to overcome the barrier associated with fusing the viral and host membranes. Although the structural intermediates of this fusion process are attractive targets for drug development, no detailed high-resolution structural information or quantitative thermodynamic characterization are available. By measuring the dynamic equilibrium between the lipid-bound intermediate and the post-fusion six-helical bundle (6HB) states of the gp41 ectodomain in the presence of bilayer membrane mimetics, we derived both the reaction kinetics and energies associated with these two states by solution NMR spectroscopy. At equilibrium, an exchange time constant of about 12 seconds at 38 °C is observed, and the post-fusion conformation is energetically more stable than the lipid-bound state by 3.4 kcal mol-1. The temperature dependence of the kinetics indicates that the folding occurs through a high-energy transition state which may resemble a 5HB structure. The energetics and kinetics of gp41 folding in the context of membrane bilayers provide a molecular basis for an improved understanding of viral membrane fusion.

Small heat-shock proteins (sHSPs) are ubiquitously expressed molecular chaperones present in all kingdoms of life that inhibit protein misfolding and aggregation. Despite their importance in proteostasis, the structure-function relationships of sHSPs remain elusive. Human sHSPs are characterised by a central, highly conserved α-crystallin domain (ACD) and variable-length N- and C-terminal regions. The ACD forms antiparallel homodimers via an extended β-strand, creating a shared β-sheet at the dimer interface. The N- and C-terminal regions mediate formation of higher order oligomers that are thought to act as storage forms for chaperone-active dimers. We investigated the interactions of the ACD of two human sHSPs, αB-crystallin (αB-C) and Hsp27, with apolipoprotein C-II amyloid fibrils using analytical ultracentrifugation and nuclear magnetic resonance spectroscopy. The ACD was found to interact transiently with amyloid fibrils to inhibit fibril elongation and naturally occurring fibril end-to-end joining. This interaction was sensitive to the concentration of fibril ends indicating a 'fibril-capping' interaction. Furthermore, resonances arising from the ACD monomer were attenuated to a greater extent than those of the ACD dimer in the presence of fibrils, suggesting that the monomer may bind fibrils. This hypothesis was supported by mutagenesis studies in which disulfide cross-linked ACD dimers formed by both αB-C and Hsp27 were less effective at inhibiting amyloid fibril elongation and fibril end-to-end joining than ACD constructs lacking disulfide cross-linking. Our results indicate that sHSP monomers inhibit amyloid fibril elongation, highlighting the importance of the dynamic oligomeric nature of sHSPs for client binding.

Viruses of the sobemovirus genus are plant viruses, most of which generate very important agricultural and financial losses. Among them, the rice yellow mottle virus (RYMV) is one of the most damaging pathogens devastating rice fields in Africa. RYMV infectivity and propagation rely on its protein P1, identified as a key movement and potential long-distance RNA silencing suppressor. Here we describe P1's complete 3D structure and dynamics obtained by an integrative approach combining X-Ray crystallography and NMR spectroscopy. We show that P1 is organized in two semi-independent and topologically unrelated domains, each harboring an original zinc finger. The two domains exhibit different affinities for zinc and sensitivities to oxidoreduction conditions, making the C-terminal P1 region a potential labile sensor of the plant redox status. An additional level of regulation resides on the capacity of P1 to oligomerize through its N-terminal domain. Coupling P1 structure information with site-directed mutagenesis and plant functional assays, we identified key residues in each zinc domain essential for infectivity and spread in rice tissues. Altogether, our results provide the first complete structure of a sobemoviral P1 movement protein and highlight structural and dynamical properties that may serve RYMV functions to infect and invade its host plant.

New variants of the severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2) emerged and spread rapidly all over the world, which strongly supports the need for pharmacological options to complement vaccine strategies. Main protease (Mpro or 3CLpro) is a critical enzyme in the life cycle of SARS-CoV-2 and appears to be highly conserved among different genera of coronaviruses, making it an ideal target for the development of drugs with broad-spectrum property. PF-07304814 developed by Pfizer is an intravenously administered inhibitor targeting SARS-CoV-2 Mpro. Here we showed that PF-07304814 displays broad-spectrum inhibitory activity against Mpros from multiple coronaviruses. Crystal structures of Mpros of SARS-CoV-2, SARS-CoV, MERS-CoV, and HCoV-NL63 bound to the inhibitor PF-07304814 revealed a conserved ligand-binding site, providing new insights into the mechanism of inhibition of viral replication. A detailed analysis of these crystal structures complemented by comprehensive comparison defined the key structural determinants essential for inhibition and illustrated the binding mode of action of Mpros from different coronaviruses. In view of the importance of Mpro for the medications of SARS-CoV-2 infection, insights derived from the present study should accelerate the design of pan-coronaviral main protease inhibitors that are safer and more effective.

Nuclear receptors constitute one of the largest families of transcription factors that regulate genes in metazoans in response to small molecule ligands. Many receptors harbor two transactivation domains, one at each end of the protein sequence. Whereas the molecular mechanisms of transactivation mediated by the ligand-binding domain at the C-terminus of the protein are generally well established, the mechanism involving the N-terminal domain called activation function 1 (AF1) has remained elusive. Previous studies implicated the AF1 domain as a significant contributor towards the overall transcriptional activity of the NR4A family of nuclear receptors and suggested that the steroid receptor coactivators (SRCs) play an important role in this process. Here we show that a short segment within the AF1 domain of the NR4A receptor Nurr1 can directly engage with the SRC1 PAS-B domain. We also show that this segment forms a helix upon binding to a largely hydrophobic groove on PAS-B, overlapping with the surface engaged by the STAT6 transcription factor, suggesting that this mode of recruitment could be shared by diverse transcription factors including other nuclear receptors.

Microbiology has long studied the ways in which subtle genetic differences between closely related microbial strains can have profound impacts on their phenotypes and those of their surrounding environments and communities. Despite the growth in high-throughput microbial community profiling, however, such strain-level differences remain challenging to detect. Once detected, few quantitative approaches have been well-validated for associating strain variants from microbial communities with phenotypes of interest, such as medication usage, treatment efficacy, host environment, or health. First, the term "strain" itself is not used consistently when defining a highly-resolved taxonomic or genomic unit from within a microbial community. Second, computational methods for identifying such strains directly from shotgun metagenomics are difficult, with several possible reference- and assembly-based approaches available, each with different sensitivity/specificity tradeoffs. Finally, statistical challenges exist in using any of the resulting strain profiles for downstream analyses, which can include strain tracking, phylogenetic analysis, or genetic association studies. We provide an in depth discussion of recently available computational tools that can be applied for this task, as well as statistical models and gaps in performing and interpreting any of these three main types of studies using strain-resolved shotgun metagenomic profiling of microbial communities.

In higher eukaryotic cells, chromosomes are folded inside the nucleus. Recent advances in whole-genome mapping technologies have revealed the multiscale features of 3D genome organization that are intertwined with fundamental genome functions. However, DNA sequence determinants that modulate the formation of 3D genome organization remain poorly characterized. In the past few years, predicting 3D genome organization based on DNA sequence features has become an active area of research. Here, we review the recent progress in computational approaches to unraveling important sequence elements for 3D genome organization. In particular, we discuss the rapid development of machine learning-based methods that facilitate the connections between DNA sequence features and 3D genome architectures at different scales. While much progress has been made in developing predictive models for revealing important sequence features for 3D genome organization, new research is urgently needed to incorporate multi-omic data and enhance model interpretability, further advancing our understanding of gene regulation mechanisms through the lens of 3D genome organization.

Human microbiome consists of trillions of microorganisms. Microbiota can modulate the host physiology through molecule and metabolite interactions. Integrating microbiome and metabolomics data have the potential to predict different diseases more accurately. Yet, most datasets only measure microbiome data but without paired metabolome data. Here, we propose a novel integrative modeling framework, Microbiome-based Supervised Contrastive Learning Framework (MB-SupCon). MB-SupCon integrates microbiome and metabolome data to generate microbiome embeddings, which can be used to improve the prediction accuracy in datasets that only measure microbiome data. As a proof of concept, we applied MB-SupCon on 720 samples with paired 16S microbiome data and metabolomics data from patients with type 2 diabetes. MB-SupCon outperformed existing prediction methods and achieved high average prediction accuracies for insulin resistance status (84.62%), sex (78.98%), and race (80.04%). Moreover, the microbiome embeddings form separable clusters for different covariate groups in the lower-dimensional space, which enhances data visualization. We also applied MB-SupCon on a large inflammatory bowel disease study and observed similar advantages. Thus, MB-SupCon could be broadly applicable to improve microbiome prediction models in multi-omics disease studies.

Biosynthetic gene clusters (BGCs) in bacterial genomes code for important small molecules and secondary metabolites. Based on the validated BGCs and the corresponding sequences of protein family domains (Pfams), Pfam functions and clan information, we develop a deep learning method e-DeepBGC, that extends DeepBGC, for detecting the BGCs and their biosynthetic class in bacterial genomes. We show that e-DeepBGC leads to reduced false positive rates in BGC identification and an increased sensitivity in identifying BGCs compared to DeepBGC. We apply e-DeepBGC to 5,666 Ref Seq bacterial genomes and detect a total of 170, 685 BGCs with an average of 30.1 BGCs in each genome. We summarize all the predicted BGCs, their functional classes and the distributions of the BGCs in different bacterial phyla.

The recognition of protein structural folds is the starting point for protein function inference and for many structural prediction tools. We previously introduced the idea of using empirical comparisons to create a data-augmented feature space called PESS (Protein Empirical Structure Space)1 as a novel approach for protein structure prediction. Here, we extend the previous approach by generating the PESS feature space over fixed-length subsequences of query peptides, and applying a sequential neural network model, with one long short-term memory cell layer followed by a fully connected layer. Using this approach, we show that only a small group of domains as a training set is needed to achieve near state-of-the-art accuracy on fold recognition. Our method improves on the previous approach by reducing the training set required and improving the model's ability to generalize across species, which will help fold prediction for newly discovered proteins.

Recent development in inferring RNA velocity from single-cell RNA-seq opens up exciting new vista into developmental lineage and cellular dynamics. However, the estimated velocity only gives a snapshot of how the transcriptome instantaneously changes in individual cells, and it does not provide quantitative predictions and insights about the whole system. In this work, we develop RNA-ODE, a principled computational framework that extends RNA velocity to quantify systems level dynamics and improve single-cell data analysis. We model the gene expression dynamics by an ordinary differential equation (ODE) based formalism. Given a snapshot of gene expression at one time, RNA-ODE is able to predict and extrapolate the expression trajectory of each cell by solving the dynamic equations. Systematic experiments on simulations and on new data from developing brain demonstrate that RNA-ODE substantially improves many aspects of standard single-cell analysis. By leveraging temporal dynamics, RNA-ODE more accurately estimates cell state lineage and pseudo-time compared to previous state-of-the-art methods. It also infers gene regulatory networks and identifies influential genes whose expression changes can decide cell fate. We expect RNA-ODE to be a Swiss army knife that aids many facets of single-cell RNA-seq analysis.

Machine learning or deep learning models have been widely used for taxonomic classification of metagenomic sequences and many studies reported high classification accuracy. Such models are usually trained based on sequences in several training classes in hope of accurately classifying unknown sequences into these classes. However, when deploying the classification models on real testing data sets, sequences that do not belong to any of the training classes may be present and are falsely assigned to one of the training classes with high confidence. Such sequences are referred to as out-of-distribution (OOD) sequences and are ubiquitous in metagenomic studies. To address this problem, we develop a deep generative model-based method, MLR-OOD, that measures the probability of a testing sequencing belonging to OOD by the likelihood ratio of the maximum of the in-distribution (ID) class conditional likelihoods and the Markov chain likelihood of the testing sequence measuring the sequence complexity. We compose three different microbial data sets consisting of bacterial, viral, and plasmid sequences for comprehensively benchmarking OOD detection methods. We show that MLR-OOD achieves the state-of-the-art performance demonstrating the generality of MLR-OOD to various types of microbial data sets. It is also shown that MLR-OOD is robust to the GC content, which is a major confounding effect for OOD detection of genomic sequences. In conclusion, MLR-OOD will greatly reduce false positives caused by OOD sequences in metagenomic sequence classification.

MOTIVATION: Continuous emergence of new variants through appearance/accumulation/disappearance of mutations is a hallmark of many viral diseases. SARS-CoV-2 variants have particularly exerted tremendous pressure on global healthcare system owing to their life threatening and debilitating implications. The sheer plurality of variants and huge scale of genomic data have added to the challenges of tracing the mutations/variants and their relationship to infection severity (if any). RESULTS: We explored the suitability of virus-genotype guided machine-learning in infection prognosis and identification of features/mutations-of-interest. Total 199,519 outcome-traced genomes, representing 45,625 nucleotide-mutations, were employed. Among these, post data-cleaning, Low and High severity genomes were classified using an integrated model (employing virus genotype, epitopic-influence and patient-age) with consistently high ROC-AUC (Asia:0.97 ± 0.01, Europe:0.94 ± 0.01, N.America:0.92 ± 0.02, Africa:0.94 ± 0.07, S.America:0.93 ± 03). Although virus-genotype alone could enable high predictivity (0.97 ± 0.01, 0.89 ± 0.02, 0.86 ± 0.04, 0.95 ± 0.06, 0.9 ± 0.04), the performance was not found to be consistent and the models for a few geographies displayed significant improvement in predictivity when the influence of age and/or epitope was incorporated with virus-genotype (Wilcoxon p_BH < 0.05). Neither age or epitopic-influence or clade information could out-perform the integrated features. A sparse model (6 features), developed using patient-age and epitopic-influence of the mutations, performed reasonably well (>0.87 ± 0.03, 0.91 ± 0.01, 0.87 ± 0.03, 0.84 ± 0.08, 0.89 ± 0.05). High-performance models were employed for inferring the important mutations-of-interest using Shapley Additive exPlanations (SHAP). The changes in HLA interactions of the mutated epitopes of reference SARS-CoV-2 were then subsequently probed. Notably, we also describe the significance of a 'temporal-modeling approach' to benchmark the models linked with continuously evolving pathogens. We conclude that while machine learning can play a vital role in identifying relevant mutations and factors driving the severity, caution should be exercised in using the genotypic signatures for predictive prognosis.