Articles

Predicting binding specificity of T-cell receptors (TCRs) and putative antigens can help improve cancer immunotherapy. Lin et al. propose RACER, which efficiently makes use of supervised machine learning to learn important molecular interactions contributing to TCR–peptide binding.

A dynamic organ-resolved model is developed by integrating metabolic and regulatory processes in type 1 diabetes, providing a depiction of network dynamics, regulation and response to perturbations in relation to variability in insulin response.

A manifold-preserving feature selection method was developed for single-cell data analysis, which selects non-redundant features to help detect rare cell populations, design follow-up studies and create targeted panels.

This work proposes a probabilistic graphical model as a formal mathematical foundation for digital twins, and demonstrates how this model supports principled data assimilation, optimal control and end-to-end uncertainty quantification.

haploSep is a computationally efficient method to infer major haplotypes and their frequencies from multiple samples of allele frequency data, and to provide improved estimates of experimentally obtained allele frequencies.

The CARseq method allows users to assess cell type-specific differential expression using RNA-seq data from bulk tissue samples, which opens up several opportunities for re-analyzing existing RNA-seq data and designing new studies.

Physics-aware deep generative models are used to design material microstructures exhibiting tailored properties. Multi-fidelity data are used to create inexpensive yet accurate machine learning surrogates for evaluating the physics-based constraints within such design frameworks.

Stimulated emission depletion (STED) microscopy allows images to be captured at a subdiffraction resolution. Here optimal transport colocalization is proposed for analyzing macromolecule distributions in high-resolution STED images.

Developing lightweight deep neural networks, while essential for edge computing, still remains a challenge. Random sketch learning is a method that creates computationally efficient and compact networks, thus paving the way for deploying tiny machine learning (TinyML) in resource-constrained devices.

A self-consistent model that bridges electrochemistry and solid-state physics is developed to fully describe the ion and electron distribution at solid electrode/electrolyte interfaces and applied to lower interfacial resistance in batteries.

Through parametric sensitivity analysis and uncertainty quantification of the CovidSim model, a subset of this model’s parameters is identified to which the code output is most sensitive. Using these allows better and more informed decisions about proposed policies.

Analysis and interpretation of relationships across different modalities of biomedical datasets is a critical but challenging task. SiMLR is a joint dimensionality reduction algorithm that is designed to perform this task.

Spiking neural network simulations are very memory-intensive, limiting large-scale brain simulations to high-performance computer systems. Knight and Nowotny propose using procedural connectivity to substantially reduce the memory footprint of these models, such that they can run on standard GPUs.

The FDRnet method addresses several issues in cancer pathway analysis, including those related to detecting functionally homogeneous subnetworks, controlling the false discovery rate of the genes and handling the computational complexity.

Multi-fidelity graph networks learn more effective representations for materials from large data sets of low-fidelity properties, which can then be used to make accurate predictions of high-fidelity properties, such as the band gaps of ordered and disordered crystals and energies of molecules.

An approach based on soft Markov state models is capable of constructing interpretable kinetic ensembles of disordered proteins.

Computational simulations show that selection for high gene expression stability can explain the stable maintenance of obsolete phenotypic switching capabilities under natural selection.

An evolutionary-based algorithm enables modeling of complex solid-solution alloys over exponential search spaces in practical time, accelerating materials design of such high-entropy alloys.