AIDO Cell 1.0
GenBio AIDeveloper-reported Virtual Cell Benchmark 1.0 spans 31 metrics across five task families; independent reproduction remains pending.
Follow biological AI through time.
Move from early physics and deep-learning platforms to protein and genomic representations, structure prediction, generative biology, RNA, virtual cells, tissue models and integrated discovery systems.
86records17years represented
AIDO.Tissue
GenBio AIPublic AIDO.Tissue checkpoints and downstream examples are available.
GenBio-PathFM
GenBio AIReported evaluations include THUNDER, HEST and PathoROB; cross-paper comparability remains task-specific.
AlphaGenome
Google DeepMindSuccessor to Enformer; complements AlphaMissense.
BoltzGen
MIT / Boltz teamAdds a design branch to the Boltz lineage.
Borzoi
Calico / academic collaboratorsConnects DNA sequence to tissue-specific RNA readouts.
Latent-X
Latent LabsFounder co-created AlphaFold at DeepMind.
LucaOne
BioMap / collaboratorsTrained across very large species diversity.
Nicheformer
Spatial biology research teamTreats spatial context as a first-class representation.
Novae
Spatial-omics research communityAdds graph-native spatial representation to BioAtlas.
Protenix
ByteDancePart of a wave of open AF3 reimplementations (with Boltz, Chai).
RFantibody
Institute for Protein Design, UWExperimentally evaluated de novo antibody designs have been reported.
RiNALMo
University of Zagreb / A*STAR collaboratorsEvaluated on RNA structure and downstream tasks.
State / Stack
Arc InstituteTrained on 270M+ cells — largest single-cell perturbation model at release.
Tahoe-x1
Tahoe TherapeuticsThe authors report state-of-the-art performance across four disease-relevant evaluation groups; independent reproduction should be tracked separately.
TranscriptFormer
Chan Zuckerberg InitiativeSeparated from the CZI platform record in BioAtlas ontology v2.
xTrimoPGLM
BioMap / Tsinghua collaboratorsPublished in Nature Methods in 2025.
AlphaProteo
Google DeepMindReported strong binding across diverse target proteins.
Boltz-1 / Boltz-2
MIT (Barzilay & Jaakkola labs)BoltzGen extends the line toward universal binder design.
Caduceus
Cornell / Tri Dao collaboratorsExplicitly encodes a fundamental symmetry of DNA.
Cell2Sentence (C2S-Scale)
Google Research + YaleLets general LLMs 'speak' single-cell biology.
Chai-1 / Chai-2
Chai DiscoveryBacked by OpenAI Startup Fund and Thrive Capital.
CONCH
Mahmood LabAdds image-language representation to the pathology layer.
CZI Virtual Cells (rBio, TranscriptFormer)
Chan Zuckerberg InitiativeAcquired the EvolutionaryScale (ESM3) team in 2025.
ESM3
EvolutionaryScale (now CZ Biohub)Generated the novel esmGFP fluorescent protein.
Evo / Evo 2
Arc Institute + Stanford + NVIDIATrained on 9.3T DNA base pairs (Evo 2).
GROVER
Poetsch Lab / collaboratorsIncluded in head-to-head genomic-FM benchmarking.
H-Optimus
BioptimusBioptimus reports training across more than one million slides for current H-Optimus releases.
HelixFold3
PaddleHelix / BaiduTechnical report compares the model with AlphaFold3-class tasks.
Prov-GigaPath
Microsoft Research / ProvidenceModels local morphology and whole-slide context.
SaProt
Westlake / Zhejiang collaboratorsCombines sequence and structural alphabets in the input representation.
SCimilarity
GenentechPublished in Nature in 2024.
TITAN
Mahmood Lab / collaboratorsPretrained on hundreds of thousands of whole-slide images in the published technical report.
UNI / UNI2
Mahmood Lab / Brigham and Women’s HospitalRepresents tissue morphology as reusable image embeddings.
UNI-RNA
Academic RNA foundation-model collaboratorsReviewed as a major RNA foundation-model family in 2026 surveys.
Virchow / Virchow2
Paige / Microsoft Research collaboratorsA prominent comparator in pathology foundation-model studies.
Xaira Therapeutics
Xaira TherapeuticsOne of the largest biotech launches ever (>$1B).
AlphaMissense
Google DeepMindClassified ~71M possible human missense variants.
Ankh
Proteinea / academic collaboratorsIncluded in modern protein-LM comparisons.
Chroma
Generate BiomedicinesBacked by Flagship Pioneering.
DNABERT-2
Multi-institution research teamA major baseline in current genomic-FM benchmarking.
Geneformer
Broad InstituteTrained on ~30M single-cell transcriptomes.
HyenaDNA
Stanford (Hazy Research)Architectural ancestor of Evo's StripedHyena.
Nucleotide Transformer
InstaDeep (BioNTech)InstaDeep was acquired by BioNTech.
NVIDIA BioNeMo
NVIDIACo-developed Evo 2 with the Arc Institute.
RFdiffusion
Institute for Protein Design, UWRFdiffusion All-Atom extends to small-molecule binders.
scFoundation
Tsinghua / BioMapTrained on ~50M single cells.
scGPT
University of Toronto (Bo Wang Lab)Pretrained on 33M+ human cells.
Universal Cell Embeddings (UCE)
Stanford / collaboratorsA major comparator in modern single-cell foundation-model evaluations.
DiffDock
MIT (Barzilay & Jaakkola labs)Open-sourced and integrated into BioNeMo.
ESM-2
Meta AI (FAIR)Provides the language-model representation used by ESMFold.
ESMFold
Meta AI (FAIR)~60x faster than MSA-based methods.
IgLM / AntiBERTy
Johns Hopkins (Gray Lab)Underpins the IgFold antibody structure predictor.
MolMIM
NVIDIAAvailable through NVIDIA BioNeMo/NIM workflows.
OpenFold
OpenFold Consortium / ColumbiaRevealed how folding models learn during training.
Profluent (ProGen3 / OpenCRISPR)
Profluent BioOpenCRISPR-1 was AI-designed and open-sourced.
ProteinMPNN / LigandMPNN
Institute for Protein Design, UWThe standard partner to RFdiffusion.
RNA-FM
ML4Bio / academic collaboratorsThe project reports pretraining on more than 23 million non-coding RNA sequences.
Uni-Mol / Uni-Mol2
DP Technology / DeepModelingUni-Mol2 scales the family to large 3D molecular pretraining corpora.
Cradle Bio
CradleFounder previously led products at Google.
Enformer
Google DeepMindPredecessor to AlphaGenome.
Isomorphic Labs (IsoDDE)
Isomorphic Labs (Alphabet)IsoDDE more than doubles AF3 accuracy on hardest targets.
MegaMolBART
NVIDIADistributed through NVIDIA BioNeMo.
ProtT5
RostlabA common protein-embedding benchmark baseline.
RoseTTAFold / All-Atom
Institute for Protein Design, UWDavid Baker won the 2024 Nobel Prize in Chemistry.
Valence Labs (MolGPS / LOWE)
Recursion PharmaceuticalsMolGPS scaled molecular graph models to new sizes.
AlphaFold 2 / 3
Google DeepMindAlphaFold DB holds 200M+ predicted structures, free for all.
Nabla Bio — JAM
Nabla BioTargets multi-pass membrane proteins directly.
NeuralPLexer / Enchant
Iambic TherapeuticsSpun out of Caltech quantum-chemistry research.
ProGen / ProGen2 / ProGen3
Salesforce Research → ProfluentProGen lysozymes were experimentally active.
Basecamp Research (EDEN / BaseFold)
Basecamp ResearchCollects samples from extreme natural environments worldwide.
BigHat Biosciences — Milliner
BigHat BiosciencesCo-founder built Google's DeepVariant.
GEMS
Genesis TherapeuticsFounded from Stanford ML + chemistry research.
Generate:Biomedicines
Generate:BiomedicinesBuilt on the Chroma generative model.
insitro
insitroFounder Daphne Koller is a Stanford ML pioneer and Coursera co-founder.
Owkin
OwkinPioneered federated learning across hospital networks.
BigRNA
Deep GenomicsFounder Brendan Frey is a deep-learning pioneer.
Chemistry42 / Pharma.AI
Insilico MedicineINS018_055 was the first drug with AI-discovered target AND molecule in Phase II.
BenevolentAI
BenevolentAIFlagged baricitinib for COVID-19; later trial-validated.
Recursion (LOWE + Phenomics)
Recursion PharmaceuticalsCompleted the acquisition of Exscientia on 20 November 2024.
Exscientia
Recursion PharmaceuticalsAmong the first with AI-designed molecules in trials.
LabGenius — EVA
LabGeniusRobotics-driven design–test loop.
Numerion AI Chemistry Platform
Numerion LabsFormerly known as Atomwise.
Absci — Integrated Drug Creation
AbsciPublicly traded AI-biotech.
Ginkgo Bioworks (Datapoints / AI Models)
Ginkgo BioworksOne of the largest cell-engineering foundries.
Schrödinger Platform
Schrödinger, Inc.Publicly traded; software used industry-wide.
The year reflects the introduction year recorded in BioAtlas. Milestones are curated summaries and should be checked against the linked primary source.