Integrating machine learning (ML) models with selected features improved the prediction accuracy of diabetic kidney disease compared to traditional logistic regression models using established features, revealing potential new markers and supporting the use of ML techniques in epidemiological studies.

Combined analyses of transcriptome and chromatin accessibility elucidated the mechanisms underlying cancer cell lines response to antitumor candidates and provided a versatile perturbation-informed basal signature able to predict drug sensitivity.

Recurrent neural network models enable prediction and design of health-relevant metabolite dynamics in synthetic human gut communities.

Orthologous proteins display different mutational robustness that can be leveraged to improve prediction of functional sequences.

A smart combination of machine learning and high-throughput data generation of bacterial population dynamics successfully leads to an intriguing finding of the differentiation in decision-making components for bacterial growth.

Shifts in pH that result from metabolic interactions between members of the Drosophila gut microbiota were sufficient to modulate Lactobacillus plantarum tolerance to the antibiotics rifampin and erythromycin.

The ubiquitous second messenger cyclic AMP, produced by Rv3645/MacE, controls long-chain fatty acid metabolism and intrinsic multidrug resistance in M. tuberculosis.

Modeling the dynamics of immunological synapses informs pharmacodynamics and treatment resistance to bispecific T cell engagers.

Machine learning models effectively predict the risk of post-stroke epilepsy using extensive clinical data, offering new insights for improving patient management in clinical neurology.

A method to assess the risk of self-sustained HIV transmission in heterosexuals from phylogenetic and epidemiological data is developed and, when applied to the Swiss HIV epidemic, shows that this risk is negligibly small for Switzerland.