Genome neighborhood networks provide an efficient large-scale approach to mine metabolic pathway context for entire families of proteins and enzymes.

NetTCR 2.2 demonstrates advances in predicting MHC-peptide-TCR interactions through integration of pan- and peptide-specific training strategies, loss-scaling and sequence similarity scoring.

A new perception of the organization of T-cell receptor repertoires in mice and humans, based on high-throughput sequencing and CDR3 sequence similarity, indicates hubs of cross-species public sequences forming evolutionary conserved 'foci of attention' of T cell immunity.

A bioinformatics pipeline has annotated over 2,500 complete genomes of circular DNA viruses belonging to dozens of established and emerging viral families.

Orthologous proteins that continuously maintain the same molecular function do not usually diverge beyond a certain level of sequence and structural similarity.

A bioinformatics approach shows how to reduce noise in single-cell TCR-pMHC specificity data while retaining sensitivity toward cross-binding events to facilitate investigation of the rules governing TCR-pMHC binding.

Evidence for multiple brain systems for sequence processing involving statistical inferences at multiple scales.

Humans memorize structured sound sequences using a language-of-thought compression algorithm.

Computational reaction representations and profile hidden Markov model searches of metagenomics databases can be harnessed to accurately predict bacterial species and enzyme sequences responsible for biotransformations in the human gut microbiome.

The analysis of mass-spectrometry data for all possible open reading frames reveals protein evidence for evolutionarily young, unannotated proteins with distinct characters.