2026-04-06

Top 40

BACKFILLED Code and industry signals are partial for backfilled days.

Single-cell

Learned models convert single-cell and spatial omics into interpretable predictions

High-dimensional single-cell and spatial omics data now feed learned frameworks that yield interpretable and clinically oriented predictions.

Structural bio & protein design

Knowledge-based and graph-network scores for protein-ligand affinity and pose

Recent work pairs direction-aware statistical potentials and separated affinity and pose graph networks with a 178,263-complex curated PDB dataset.

Single-cell

Background baselines recover TF-specific signatures from pooled single-cell screens

Re-analysis shows explicit baselines recover TF-specific signatures from pooled single-cell screens with missing controls.

Tools & infrastructure

Formalized analysis tools refine metagenomic, fungal, and proteomic abundance estimates

Analysis decisions in metagenomics, mycobiome profiling, and proteomics are being packaged into reproducible, benchmarked software procedures.

Tools & infrastructure

Tools address identifier mapping, sequencing QC, and reproducible workflow management

Recent preprints and a workflow paper rebuild core analysis steps for speed, scope, and reproducibility across omics.

6 Refining variant calls by modelling context-specific error and reference structureGenomics 7 Protein language models predict previously unknown prokaryotic defense systemsML / AI methods 8 Reusable representations of count, missing, and batch-affected biological measurementsML / AI methods 9 Structure predictors and binding models probe interfaces and conformational statesStructural bio & protein design 10 Quantitative methods probe neural oscillations, connectivity, and manifold geometrySystems biology 11 Machine learning models infer origin, selection, and fitness from genomic dataGenomics 12 ML models decode antibiotic action, microbial states, viral lineages and enzyme classesML / AI methods 13 MD ensembles and replica-exchange sampling map folding and binding pathwaysStructural bio & protein design 14 Generative models infer unobserved single-cell transitions from snapshot measurementsSingle-cell 15 Physical reweighting and feature filtering refine protein conformational ensemblesStructural bio & protein design 16 Tokenization and uncertainty limit biological foundation model representationsML / AI methods 17 Peptide and ion models validated beyond their training regimes against external reference…Structural bio & protein design 18 Population-scale genomes expose structure, hidden variation, and imputation limitsGenomics 19 Integrated chromatin maps link enhancer-promoter wiring and histone marks to cell stateEpigenomics 20 Localization-aware and spatially guided methods resolve tumor molecular heterogeneityMetabolomics