Top 40
View all ›BACKFILLED Code and industry signals are partial for backfilled days.
Single-cell
Learned models integrate spatial and single-cell omics across modalities and tissues
Spatial and single-cell profiling now spans morphology, multiomic, and perturbation data, demanding scalable and robust inference methods.
Structural bio & protein design
Learned models now predict and generate drug-like molecules, peptides, and binders
Recent preprints pair generative design with calibrated uncertainty and foundation-model embeddings for drug-target, ADMET, and siRNA prediction.
ML / AI methods
Prior biological knowledge built into transcriptomic deep learning architectures
Transcriptomic models increasingly encode gene regulatory priors, LLM gene knowledge, and pathway topology directly in their architectures.
Structural bio & protein design
Learned and enhanced-sampling models are mapping protein conformational ensembles
Machine-learned coarse-grained models and benchmarked enhanced-sampling workflows now target transferable, Boltzmann-weighted protein ensemble estimates.
ML / AI methods
Benchmarks test whether biological response models generalize to unseen data
Leakage-audited and calibrated benchmarks now quantify when drug-response and perturbation models transfer to new patients and contexts.