2025-12-28

Top 40

BACKFILLED Code and industry signals are partial for backfilled days.

Single-cell

Structure-aware models learn spatial and single-cell omics representations

Recent preprints extend integration, batch correction, and per-cell quality control to imaging-based spatial and single-cell multi-omics data.

Structural bio & protein design

Pretrained sequence models predict stability, resistance and protein interactions

Recent preprints test pretrained protein and genomic models on task-level endpoints such as ΔTm stability and antimicrobial resistance.

Tools & infrastructure

Large-scale genome and metagenome mining links sequence content to microbial function

Expanding genome and metagenome collections now allow trait-specific functions to be assigned from broad microbial sampling.

ML / AI methods

Attention graph networks over biomedical graphs predict molecular and genetic outcomes

Heterogeneous biomedical graphs are now paired with attention, ensemble, and multi-view architectures for diverse prediction tasks.

ML / AI methods

Learned visual models for cell phenotyping and tissue classification in microscopy

Pathology foundation-model embeddings and label-efficient classifiers are now being applied to H&E and multiplexed immunofluorescence cell and tissue analysis.

6 Genome rearrangements and regulatory co-option shape cell identity and lineage traits↑ 1.9×Genomics 7 LLM agents link literature retrieval to gene and protein annotation↑ 1.8×ML / AI methods 8 Designed mini-protein binders reach nanomolar affinity to viral receptor-binding domains↑ 1.6×Structural bio & protein design 9 Tools harmonize crosslinking MS, metabolomics and RNA-seq outputs for downstream analysis↑ 1.6×Tools & infrastructure 10 Structure-based models for protein-ligand affinity, screening, and property predictionML / AI methods 11 Analysis toolkits for single-cell DNA, lentiviral integration, and mutational signaturesGenomics 12 Experimental restraints and force fields refine RNA and peptide conformational ensembles↑ 1.6×Structural bio & protein design 13 ML models predict immunogenicity for designed binders and neoepitopesML / AI methods 14 Polymer models link epigenetic states to chromatin mechanics and Hox collinearityEpigenomics 15 LLMs and deep recommenders extend drug interaction and target prediction modelsML / AI methods 16 Polarizable simulation and AF3 advance RNA/DNA binding and ssDNA structure predictionStructural bio & protein design 17 Models recover temporal structure from static or sparsely timed cell measurementsSingle-cell 18 Intrinsic noise limits timing and adaptation in gene regulatory circuitsSystems biology 19 ModkitOpt tunes nanopore direct RNA modification calling for higher precisionSingle-cell 20 Generative and CNN models now design cell-type-specific regulatory elements and enhancersML / AI methods

Show 10 more

21 Machine learning predicts pathogenicity of mitochondrial noncoding and missense variantsGenomics 22 Haplotype-resolved assemblies tie plant genome structure to resistance traitsGenomics 23 Faster LD score, local heritability, and ancestry inference methods for trait geneticsGenomics 24 Deep learning models for DNA, mRNA and tRNA sequence prediction and designGenomics 25 Benchmarking omics methods against simulated references and disease biomarker dataML / AI methods 26 Simulations map residue-level determinants of amyloid aggregation and its inhibitionStructural bio & protein design 27 Explicit computational models of cell, tissue, and embryonic systems against experimental…Systems biology 28 Pretrained models constrained by geometry and chemistry drive generative molecular designStructural bio & protein design 29 Learned models map microbial biosynthesis from literature, genomes and pathway structuresGenomics 30 Linking genomic factors and causal instruments to multi-omics molecular networksSystems biology