Schrödinger Platform
Physics-based simulation, now fused with machine learning.
Evidence1/7 evidence fields documented
AccessProprietary
ComputeCPU/GPU / commercial platform
BenchmarkNot yet curated
Virtual Cell and Perturbation AI Models
Virtual-cell predictions are highly sensitive to cell type, tissue, dose, time and experimental context. BioAtlas separates representation learning from perturbation prediction and experimental validation.
Curated model passports
17 relevant BioAtlas records.
Open evidence passport →
Geneformer
A transfer-learning foundation model of gene networks.
Evidence2/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkNot yet curated
BigRNA
An RNA foundation model for oligonucleotide therapeutics.
Evidence1/7 evidence fields documented
AccessProprietary
ComputeVendor managed
BenchmarkNot yet curated
State / Stack
The first virtual-cell model predicting how cells react to drugs.
Evidence4/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkPerturbation prediction datasets
CZI Virtual Cells (rBio, TranscriptFormer)
A moonshot to build AI models of every human cell.
Evidence1/7 evidence fields documented
AccessOpen source
ComputePlatform dependent
BenchmarkNot yet curated
scGPT
A GPT for the language of single cells.
Evidence4/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkSingle-cell downstream tasks
Cell2Sentence (C2S-Scale)
Turning cells into 'sentences' that an LLM can reason over.
Evidence2/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkNot yet curated
scFoundation
A 100M-parameter model of ~20,000 genes at once.
Evidence2/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkNot yet curated
Recursion (LOWE + Phenomics)
An industrial-scale phenomics + ML drug-discovery engine.
Evidence1/7 evidence fields documented
AccessLimited open access
ComputePlatform dependent
BenchmarkNot yet curated
Xaira Therapeutics
A $1B+ launch fusing Baker-lab design with drug development.
Evidence1/7 evidence fields documented
AccessProprietary
ComputeVendor managed
BenchmarkNot yet curated
insitro
Machine-learning-driven biology for drug discovery.
Evidence1/7 evidence fields documented
AccessProprietary
ComputeVendor managed
BenchmarkNot yet curated
Borzoi
DNA-to-RNA regulatory modelling across cell and tissue contexts.
Evidence4/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkFunctional-genomics and RNA-seq evaluations
Universal Cell Embeddings (UCE)
Cell embeddings designed to transfer across tissues, datasets and species.
Evidence2/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkCross-dataset single-cell evaluations
TranscriptFormer
Cross-species transcriptomic foundation modelling for cell representation and generation.
Evidence1/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkNot yet curated
SCimilarity
A cell-atlas foundation model for scalable similarity search across expression states.
Evidence3/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkNot yet curated
Nicheformer
Foundation modelling for cells with tissue-neighborhood and spatial context.
Evidence1/7 evidence fields documented
AccessLimited open access
ComputeGPU recommended
BenchmarkNot yet curated
Novae
Graph-native representation learning for spatial transcriptomics domains and niches.
Evidence1/7 evidence fields documented
AccessOpen source
ComputeGPU recommended
BenchmarkNot yet curated
How to interpret this topic
Choose by context of use.
Inputs matter
Sequence, structure, ligand, assay and single-cell inputs imply different data-quality and preprocessing assumptions.
Outputs are not interchangeable
A predicted pose, confidence score, affinity estimate, generated sequence and perturbation profile support different decisions.
Benchmarks are protocol-bound
Claims should only be compared when task, dataset, split, metric and evaluation protocol genuinely align.
Validation remains external
BioAtlas records evidence boundaries; prospective scientific and experimental validation is still required for consequential use.