2026-04-03

Top 40

BACKFILLED Code and industry signals are partial for backfilled days.

ML / AI methods

Protein language models combined with structural and experimental constraints

Early-April 2026 preprints and publications test pretrained protein models against structural data, experimental constraints, and benchmark baselines.

Genomics

Integrative genomics resolves enhancer-promoter wiring and chromatin regulatory states

Single-cell four-omics and HiChIP-based TF footprinting methods now jointly link genome conformation, chromatin marks, accessibility and expression.

Structural bio & protein design

Sampled ensembles and pathways replace static structures in protein binding and folding

Recent work applies MD sampling to binding and folding landscapes across nanobodies, lasso peptides, scFvs, and aptamers.

Single-cell

Explicit confound modeling in single-cell, proteomic and enrichment analyses

Pooled single-cell screens, RNA-protein discordance, and marker detection are now analyzed with explicit models of the confounding layers that limit…

Tools & infrastructure

Automated tools for metagenomic binning, abundance estimation, and transductomics

Repurposed human sequencing data and packaged, rule-based analysis decisions are making metagenomic and mobile-element profiling more reproducible.

6 Modelling error and reference context to refine somatic and mosaic variant detectionGenomics 7 Learned representations compress and align omics data for interpretable predictionML / AI methods 8 Generative models infer unobserved cell-state transitions from single-cell dataSingle-cell 9 New bioinformatics tools pursue speed, structural annotation, and reproducible workflowsTools & infrastructure 10 Computational frameworks mine bacterial dark proteomes and antibiotic responsesTools & infrastructure 11 New models and estimators probe neural connectivity and dynamicsSystems biology 12 Method choices reshape inferred protein conformational states from MD and AI modelsStructural bio & protein design 13 Knowledge-based and graph-learned potentials score protein-ligand and metal bindingStructural bio & protein design 14 Protein language models predict unknown bacterial defense and virus-host interactionsML / AI methods 15 Deep learning models link DNA sequence to cell-type regulatory activityML / AI methods 16 k-mer and machine learning tools advance homology, phylogenomics, alignment analysisGenomics 17 MD simulations link conserved waters and ligand binding to receptor and enzyme variationStructural bio & protein design 18 Reference content and modeling choices govern genotype and genomic prediction accuracyGenomics 19 Tools for scaling access to large biological datasets and integrating spatial omics…Tools & infrastructure 20 Haplotype-resolved buckwheat genome and reference-free pangenome across AurantioideaeGenomics