Top 40
View all › Rank 1
ML / AI methods
Single-cell and DNA models face simple baselines, attribution audits and retrieval tests
Benchmarks test single-cell and DNA models against simple baselines and replicates, while agentic and SQL interfaces aim for grounded, checkable outputs.
7 preprints · 2 published · ↑ 3.2× vs 4-wk avg
Rank 2 Single-cell
Spatial methods separate composition, state and niche; trajectory models add mechanism
Spatial tools now attribute expression to abundance, state, niche or single molecules; trajectory tools add GRNs, lineage barcodes and plausibility metrics.
16 preprints · 1 published · ↑ 2.4× vs 4-wk avg
Rank 3 Transcriptomics
Molecular profiles are being mapped to patient toxicity, survival, and drug response
irAEGIS and MANTIS predict individual toxicity and survival, while CALM, DiseaseCIO and RNAactDrug 2.0 link omics data to clinical and drug outcomes.
3 preprints · 8 published · ↑ 2.2× vs 4-wk avg
Rank 4 Single-cell
Perturbation response models add cross-context transfer and regulatory-path attribution
New preprints combine zero-shot, in-context and pretrained response prediction with inference of the regulatory paths and pathways perturbations act through.
11 preprints · 1 published · ↑ 2.1× vs 4-wk avg
Rank 5 ML / AI methods
Histology and host transcriptomes used to infer unmeasured spatial molecular layers
New preprints predict spatial protein, transcriptome and pathogen maps from histology or host RNA, alongside missing-modality survival models.
7 preprints · 2 published · ↑ 2.0× vs 4-wk avg
Rank 6 Frozen single-cell foundation model embeddings are reused for cross-modality completion…↑ 2.0×Single-cellRank 7 Curated portals and workflows consolidate spatial and single-cell data into standardized,…↑ 1.8×Tools & infrastructureRank 8 Sequence data drive microbial classification, from DNA language models to strain typing…↑ 1.9×GenomicsRank 9 Biomedical LLM agents are being constrained by harnesses, memory and audits rather than…↑ 1.9×ML / AI methodsRank 10 Coupled mesoscale and polymer models link driving processes to chromatin and condensate…↑ 1.8×EpigenomicsRank 11 Oligo probe and pool design consolidated into single installable software packages↑ 1.7×Tools & infrastructureRank 12 Protocol-matched benchmarks and foundation encoders reshape ML prediction of…↑ 1.7×ML / AI methodsRank 13 Methods operate on raw reads and pileups while profiling memory and throughput limits↑ 1.7×GenomicsRank 14 Curated microbiome catalogs and long-read tools link microbial genomes to environmental…↑ 1.7×Tools & infrastructureRank 15 Transfer and domain adaptation carry cancer transcriptomic models across cohorts, cell…↑ 1.7×ML / AI methodsRank 16 Sequence-based models predict single-molecule chromatin states and motif usage across…↑ 1.8×EpigenomicsRank 17 New quantifiers cut isoform ambiguity via k-mer specificity, self-benchmarking and…TranscriptomicsRank 18 Pangenome graphs resolve structural variation and gene content across microbial and yeast…GenomicsRank 19 Whole-genome doubling is being resolved as a subclonal, timed event rather than a binary…GenomicsRank 20 Network-centric tools couple topology, annotation, and visualization with omics data in R…↑ 1.6×Tools & infrastructure
Show 20 moreShow less
Rank 21 Interpretable models resolve tissue- and cell-context-specific disease signaturesML / AI methodsRank 22 Curated databases pair experimental binding and specificity data with structuresTools & infrastructureRank 23 Deep and unsupervised models prioritize noncoding variants by regulatory impactEpigenomicsRank 24 Assembly-level genomes and annotation tools span mosses, annelids and algal relativesGenomicsRank 25 Inferring glioblastoma invasion and proliferation parameters from imaging-derived modelsSystems biologyRank 26 Frameworks and resources standardize scoring of diverse human variant classesTools & infrastructureRank 27 Computational tools test and accelerate RNA structure prediction and validationStructural bio & protein designRank 28 Model-guided design of proteins and molecular sequences tested against measured outcomesStructural bio & protein designRank 29 Web and database tools link structure, bioactivity and genome data to candidate peptidesTools & infrastructureRank 30 Statistical inference methods for microbiome differential abundance and feature selectionML / AI methodsRank 31 Testing transcriptomic inferences against confounders and independent readoutsEpigenomicsRank 32 Scoring predicted complexes, epitopes and RNA-protein contactsStructural bio & protein designRank 33 Large cohorts resolve non-coding and population-specific genetic architecture across…GenomicsRank 34 Pretrained sequence models target DNA, RNA, and protein prediction tasksML / AI methodsRank 35 Reference-matched assemblies enable cistrome and methylation-variant analysis in cropsGenomicsRank 36 Computational pipelines for cryo-EM reconstruction and ligand-density analysisStructural bio & protein designRank 37 Explicit references for judging biological model outputs in genome folding and cryo-EMEpigenomicsRank 38 Generative models sample protein conformational ensembles constrained by experimental dataStructural bio & protein designRank 39 Benchmarks expose shared de novo peptide errors and test fragment-ion priors in MS/MSProteomicsRank 40 Graph and network models link metabolite structures, microbes, and protein interactionsMetabolomics