2025-11-03

Top 40

PARTIAL Some sources or summaries were unavailable for this run.

BACKFILLED Code and industry signals are partial for backfilled days.

ML / AI methods

Models fuse multi-omics, imaging and clinical inputs for prediction and subtyping

Heterogeneous omics, pathology and clinical data are increasingly combined in graph, neural and factorization models for prediction and subtyping.

Structural bio & protein design

Learned and physics-based models predict protein binding affinity, sites and partners

Sequence-plus-structure graph and language-model architectures are now applied to partner-specific prediction, alongside benchmarks exposing scorer bias.

Single-cell

Computational methods reconstruct single-cell and spatial profiles from incomplete data

Recent publications benchmark LLM and reference-based cell annotation alongside new spatial and cryo-ET localization methods for sparse or indirect…

ML / AI methods

Protein language model embeddings combined with structure for function prediction

Pretrained protein language model embeddings recur as inputs across function, interaction, structure, and allergenicity prediction in recent publications.

Single-cell

Temporal ordering of spatial transcriptomic microenvironments and cell fate trajectories

Sequential spatial slides and multi-timepoint single-cell data now support explicit modeling of temporal ordering in tissue and cell-fate inference.

6 Pretrained and attention-based networks predict RNA, DNA and protein functional sites↑ 1.5×ML / AI methods 7 Computational methods address noise and bias in sequencing-based genomic profiling↑ 1.5×Epigenomics 8 Benchmarks and packaged pipelines reshape plasmid, phage, and virome sequence analysis↑ 1.5×Tools & infrastructure 9 Chromosome-level assemblies of annelid, cricket and truffle genomes with new analysis…Genomics 10 Multimodal deep learning and benchmarking reshape computational drug predictionML / AI methods 11 Benchmarks and tools refine variant calling, CNV inference, and splicing analysisGenomics 12 New software spans genomic tokenization, LLM metadata curation, and variant analysisTools & infrastructure 13 Deep learning models predict methylation, TF binding, and variant effects from sequenceEpigenomics 14 Evaluating how well predictors recover RNA-containing biomolecular structuresStructural bio & protein design 15 Deep learning models interpret noncoding and missense variant effects across genomesEpigenomics 16 Transformers extend to antibody, Alzheimer's, and proteomic risk predictionML / AI methods 17 Machine learning and simulation methods target ligand kinetics, peptides and aptamersStructural bio & protein design 18 Open software packages integrate heterogeneous omics layers via network-level modelingTools & infrastructure 19 Deep learning models fuse multimodal drug data for synergy and interaction predictionML / AI methods 20 Web servers and containers package single-cell, spatial, breeding and metabolic methodsTools & infrastructure

Show 10 moreShow less

21 Faster, leaner algorithms for alignment, motif scoring, and genome size estimationTools & infrastructure 22 Post-GWAS methods adapt fine-mapping and causal gene inference to relatedness and method…Genomics 23 Learned models now generate and optimize TCR, antibody, and neoantigen sequencesML / AI methods 24 Structuring dispersed bioinformatics knowledge into linked, machine-usable resourcesTools & infrastructure 25 Inference methods recover bursting kinetics and regulatory structure in single-cell dataSingle-cell 26 Graph contrastive autoencoders and GCNs for spatial transcriptomics domain identificationTranscriptomics 27 Deep learning assigns sequencing reads and traces to discrete biological sourcesML / AI methods 28 Mechanisms ViralStructural bio & protein design 29 Surface-based and connectivity models refine tau PET and cortical MRI analysisSystems biology 30 Generative and ligand-based methods advance small-molecule and covalent ligand designStructural bio & protein design