Integration of language model and geometric deep learning enables accurate and efficient genome-scale annotation of comprehensive protein-ligand binding sites.

Protein language deep learning models can quickly and accurately translate amino acid sequences into profile hidden Markov models or a structure alphabet, dramatically improving remote homology search sensitivity without compromising space or time efficiency.

Integrating multiple protein language models using protein geometric graphs can dramatically improve the model performance for predicting the contacting residue pairs between interacting proteins.

An iterative procedure using language models allows the generation of sequences from protein families, which score similarly to natural and experimentally validated sequences, with particular promise for small families.

The recent developments in large-scale machine learning, especially with the recent Transformer models, display much potential for solving computational problems within protein biology and outcompete traditional computational methods in many recent studies and benchmarks.

A deep learning-based toolkit predicting the 3D structures of monoclonal antibodies and nanobodies outperforms other current computational methods.

A statistical model systematically associates functions of intrinsically disordered regions with sequence-distributed molecular features such as charge, residue composition, or repeat content.

A near-complete flux balance analysis model of a minimal cell demonstrates the high essentiality of its metabolic genes, agrees well with experimental essentiality data and suggests some further gene removals.

Calcium overload in cardiomyocytes disrupts sarcomere integrity, by stimulating transcriptional upregulation of the E3 ubiquitin ligase MuRF1.

For analyzing time-dependent patch-clamp or patch-clamp fluorometry data of ion channels in terms of Markovian models, the superiority of Bayesian filtering with respect to traditional deterministic approaches is demonstrated enabling more reliable quantification of the parameters.