2025-12-30

Top 40

BACKFILLED Code and industry signals are partial for backfilled days.

ML / AI methods

Learned models are applied to spatial omics, single-cell, and microscopy data

Spatial omics, single-cell profiling and microscopy datasets are now routinely analysed with transformer, autoencoder and self-supervised models.

ML / AI methods

Pretrained sequence models are repurposed for bacterial and genomic prediction tasks

Recent preprints repurpose pretrained and attention-based sequence models across bacterial genomics, RNA, and mRNA design endpoints.

Structural bio & protein design

Pretrained structure and sequence models repurposed for generative biomolecular design

Diffusion, all-atom predictors and reinforcement learning now steer biomolecular generation toward binding, stability and developability objectives.

Structural bio & protein design

Restraints and force-field tests refine conformational ensembles of flexible biomolecules

Martini3-NMR and peptide-level force field benchmarks show experimental restraints being built directly into ensemble simulations.

Single-cell

Deep learning recovers cellular dynamics and phases from untimed omics profiles

Single-cell and multi-omic data increasingly carry temporal and regulatory structure that deep generative and interpretable models can now recover.

6 Free-energy and co-folding benchmarks for protein-ligand binding and selectivityStructural bio & protein design 7 Protein AI models fail to consistently predict antibody developability propertiesML / AI methods 8 Large-scale genome and metagenome mining links sequence content to microbial functionsTools & infrastructure 9 Comparative genomics traces gene, miRNA and domain gains and losses across lineagesGenomics 10 Polymer models link epigenetic state to locus-specific chromatin mechanics and energeticsEpigenomics 11 Summary-statistic methods extend heritability and burden analysis beyond single variantsGenomics 12 Heterogeneous graph attention predicts drug effects, interactions, and binding affinityML / AI methods 13 Structure, ensemble learning and network sampling for bacterial protein predictionSystems biology 14 Intrinsic DNA sequence features direct origins, nucleosomes and enhancer identityEpigenomics 15 LLM agents mediate retrieval and evaluation over biomedical literature and datasetsML / AI methods 16 Machine learning classifiers predict cancer subtypes and outcomes from omics dataML / AI methods 17 Designed tissue-specific enhancers and in vivo single-cell CRISPR perturbation readoutsML / AI methods 18 MD simulations map residue-level determinants of amyloid aggregation and inhibitionStructural bio & protein design 19 Packaging biomolecular analysis routines into accessible web and Python workflowsTools & infrastructure 20 Intrinsic noise limits timing precision and bistability in gene regulatory networksSystems biology

Show 9 moreShow less

21 In silico fusion neoantigen discovery from pediatric cancer transcriptomesGenomics 22 Targeted primer and amplicon tools improve enrichment in viral and barcoding sequencingGenomics 23 New alignment and assembly tools target repeats, barcodes, long reads, and protein MSAGenomics 24 Neural-network state-space models and latent selection for brain activitySystems biology 25 ML generation of pathogen-targeted antimicrobial peptides with programmable propertiesML / AI methods 26 Computable models of tissue transport, signaling, and mechanics across scalesSystems biology 27 Provenance-tracked infrastructure for morphology, omics and biodiversity dataTools & infrastructure 28 Variant-aware guide design and stable multicopy constructs for genetic tool developmentTools & infrastructure 29 Computational tools sample conformations and design protein sequences in silicoStructural bio & protein design