Medium optimization is a crucial step during cell culture for biopharmaceutics and regenerative medicine; however, this step remains challenging, as both media and cells are highly complex systems. Here, we addressed this issue by employing active learning. Specifically, we introduced machine learning to cell culture experiments to optimize culture medium. The cell line HeLa-S3 and the gradient-boosting decision tree algorithm were used to find optimized media as pilot studies. To acquire the training data, cell culture was performed in a large variety of medium combinations. The cellular NAD(P)H abundance, represented as A450, was used to indicate the goodness of culture media. In active learning, regular and time-saving modes were developed using culture data at 168 h and 96 h, respectively. Both modes successfully fine-tuned 29 components to generate a medium for improved cell culture. Intriguingly, the two modes provided different predictions for the concentrations of vitamins and amino acids, and a significant decrease was commonly predicted for fetal bovine serum (FBS) compared to the commercial medium. In addition, active learning-assisted medium optimization significantly increased the cellular concentration of NAD(P)H, an active chemical with a constant abundance in living cells. Our study demonstrated the efficiency and practicality of active learning for medium optimization and provided valuable information for employing machine learning technology in cell biology experiments.

Constraint-based models (CBMs) are used to study metabolic network structure and function in organisms ranging from microbes to multicellular eukaryotes. Published CBMs are usually generic rather than context-specific, meaning that they do not capture differences in reaction activities, which, in turn, determine metabolic capabilities, between cell types, tissues, environments, or other conditions. Only a subset of a CBM's metabolic reactions and capabilities are likely to be active in any given context, and several methods have therefore been developed to extract context-specific models from generic CBMs through integration of omics data. We tested the ability of six model extraction methods (MEMs) to create functionally accurate context-specific models of Atlantic salmon using a generic CBM (SALARECON) and liver transcriptomics data from contexts differing in water salinity (life stage) and dietary lipids. Three MEMs (iMAT, INIT, and GIMME) outperformed the others in terms of functional accuracy, which we defined as the extracted models' ability to perform context-specific metabolic tasks inferred directly from the data, and one MEM (GIMME) was faster than the others. Context-specific versions of SALARECON consistently outperformed the generic version, showing that context-specific modeling better captures salmon metabolism. Thus, we demonstrate that results from human studies also hold for a non-mammalian animal and major livestock species.

Type 2 Diabetes (T2D) is often managed with metformin as the drug of choice. While it is effective overall, many patients progress to exhibit complications. Strategic drug combinations to tackle this problem would be useful. We constructed a genome-wide protein-protein interaction network capturing a global perspective of perturbations in diabetes by integrating T2D subjects' transcriptomic data. We computed a 'frequently perturbed subnetwork' in T2D that captures common perturbations across tissue types and mapped the possible effects of Metformin onto it. We then identified a set of remaining T2D perturbations and potential drug targets among them, related to oxidative stress and hypercholesterolemia. We then identified Probucol as the potential co-drug for adjunct therapy with Metformin and evaluated the efficacy of the combination in a rat model of diabetes. We find Metformin-Probucol at 5:0.5 mg/kg effective in restoring near-normal serum glucose, lipid, and cholesterol levels.

A maximum-tolerated dose (MTD) reduces the drug-sensitive cell population, though it may result in the competitive release of drug resistance. Alternative treatment strategies such as adaptive therapy (AT) or dose modulation aim to impose competitive stress on drug-resistant cell populations by maintaining a sufficient number of drug-sensitive cells. However, given the heterogeneous treatment response and tolerable tumor burden level of individual patients, determining an effective dose that can fine-tune competitive stress remains challenging. This study presents a mathematical model-driven approach that determines the plausible existence of an effective dose window (EDW) as a range of doses that conserve sufficient sensitive cells while maintaining the tumor volume below a threshold tolerable tumor volume (TTV). We use a mathematical model that explains intratumor cell competition. Analyzing the model, we derive an EDW determined by TTV and the competitive strength. By applying a fixed endpoint optimal control model, we determine the minimal dose to contain cancer at a TTV. As a proof of concept, we study the existence of EDW for a small cohort of melanoma patients by fitting the model to longitudinal tumor response data. We performed identifiability analysis, and for the patients with uniquely identifiable parameters, we deduced patient-specific EDW and minimal dose. The tumor volume for a patient could be theoretically contained at the TTV either using continuous dose or AT strategy with doses belonging to EDW. Further, we conclude that the lower bound of the EDW approximates the minimum effective dose (MED) for containing tumor volume at the TTV.

Genome-scale metabolic models (GEMs) are extensively used to simulate cell metabolism and predict cell phenotypes. GEMs can also be tailored to generate context-specific GEMs, using omics data integration approaches. To date, many integration approaches have been developed, however, each with specific pros and cons; and none of these algorithms systematically outperforms the others. The key to successful implementation of such integration algorithms lies in the optimal selection of parameters, and thresholding is a crucial component in this process. To improve the predictive accuracy of context-specific models, we introduce a new integration framework that improves the ranking of related genes and homogenizes the expression values of those gene sets using single-sample Gene Set Enrichment Analysis (ssGSEA). In this study, we coupled ssGSEA with GIMME and validated the advantages of the proposed framework to predict the ethanol formation of yeast grown in the glucose-limited chemostats, and to simulate metabolic behaviors of yeast growth in four different carbon sources. This framework enhances the predictive accuracy of GIMME which we demonstrate for predicting the yeast physiology in nutrient-limited cultures.

The exact mechanism controlling cell growth remains a grand challenge in developmental biology and regenerative medicine. The Drosophila wing disc tissue serves as an ideal biological model to study mechanisms involved in growth regulation. Most existing computational models for studying tissue growth focus specifically on either chemical signals or mechanical forces. Here we developed a multiscale chemical-mechanical model to investigate the growth regulation mechanism based on the dynamics of a morphogen gradient. By comparing the spatial distribution of dividing cells and the overall tissue shape obtained in model simulations with experimental data of the wing disc, it is shown that the size of the domain of the Dpp morphogen is critical in determining tissue size and shape. A larger tissue size with a faster growth rate and more symmetric shape can be achieved if the Dpp gradient spreads in a larger domain. Together with Dpp absorbance at the peripheral zone, the feedback regulation that downregulates Dpp receptors on the cell membrane allows for further spreading of the morphogen away from its source region, resulting in prolonged tissue growth at a more spatially homogeneous growth rate.

Multi-omics datasets are becoming of key importance to drive discovery in fundamental research as much as generating knowledge for applied biotechnology. However, the construction of such large datasets is usually time-consuming and expensive. Automation might enable to overcome these issues by streamlining workflows from sample generation to data analysis. Here, we describe the construction of a complex workflow for the generation of high-throughput microbial multi-omics datasets. The workflow comprises a custom-built platform for automated cultivation and sampling of microbes, sample preparation protocols, analytical methods for sample analysis and automated scripts for raw data processing. We demonstrate possibilities and limitations of such workflow in generating data for three biotechnologically relevant model organisms, namely Escherichia coli, Saccharomyces cerevisiae, and Pseudomonas putida.