Top 40
View all ›BACKFILLED Code and industry signals are partial for backfilled days.
Single-cell
Computational methods couple chromatin and spatial context to infer regulatory programs
New single-cell and spatial methods pair chromatin accessibility, neighborhood structure, and ligand-receptor geometry to resolve regulatory states.
Single-cell
Learned latent models recover biological structure from noisy transcriptomic measurements
Population-scale and multimodal single-cell datasets demand latent models that handle covariates and missing modalities explicitly.
Tools & infrastructure
Read-level and benchmarked pipelines for metagenomic and amplicon sequence analysis
Learned models on raw reads and unitigs and simulation-based benchmarks now expose pipeline-level performance gaps in microbiome and amplicon analysis.
Structural bio & protein design
Cofolding and learned predictors tested on antibody, kinase, and protease targets
Benchmarks of cofolding models on kinase conformations and antibody-antigen complexes show sampling and top-model ranking as the main limits.
ML / AI methods
Graph-based learned representations are routed into task-specific prediction heads
Graph encoders with contrastive, ontology, and cross-attention objectives now span protein function, drug binding, and glycan tasks.