datasetPartly relevant✓ VerifiedVC0042
Zhang J et al. · Cell · 2026-09-01
Tahoe-100M peer-reviewed (Cell 2026).
DOIPubMed
datasetPartly relevant✓ VerifiedVC0041
Zhang J et al. · bioRxiv · 2025-02-24
Tahoe-100M preprint: 100M cells, 1,100 drugs x 50 cancer cell lines.
DOI
preprint✓ VerifiedVC0138
Yu et al. · bioRxiv · 2026-08-12
PerturbLDM: Tahoe-100M-pretrained latent diffusion for conditional sc response generation across drug/dose/cell-line.
DOI
news✓ VerifiedVC0157
Biohub (with U.S. DOE Office of Science and NIH) · Biohub news (organisation announcement) · 2026-10-07
Organisation announcement (funding and data commitment, not a model or result): Biohub, the U.S. DOE Office of Science and NIH expand the Virtual Biology Initiative (announced April 2026), with a stated $1.8 billion in funding, data, computation and measurement technology for open, AI-ready biological data. Stated components: DOE more than $500M over five years; NIH to coordinate datasets and repositories from more than $500M in prior federal investment (Bio Genesis Mission); Google DeepMind, Isomorphic Labs and Meta collectively $300M; Biohub's founding $500M commitment. No data-release timeline is given.
Source
news✓ VerifiedVC0143
Arc Institute · Arc news · 2026-10-01
Arc's description of how the 2026 Challenge dataset was generated (organisation statement, not independent evaluation): CRISPRi Perturb-seq in six human cell lines, 10x Flex profiling, >10,000 perturbations and 100 non-targeting guides per line, curated to perturbations with >=80% median on-target knockdown and subsampled to 400 cells per perturbation; zero-shot task (only unperturbed cells are given); no model results reported.
Source
newsBrain tumor✓ VerifiedVC0148
aiPTO TechBio (Business Wire press release) · Business Wire · 2026-10-08
Company press release, single-party claim, pending independent verification (not evidence of model performance). The release states that DELPHAI achieved the #1 ranking on differentially expressed gene recovery in two benchmarking frameworks (Wei et al., Nat Methods 2026, DOI 10.1038/s41592-025-02980-0; Radig et al., Genome Biol 2026, DOI 10.1186/s13059-026-04104-z), describes a 'Fitness Gating' mechanism, and says the company's IPTO organoid platform covers 500+ patients across 40+ CNS tumor subtypes (company figure). No numerical results are given; training on IPTO data is described as in progress. Relates to VC0147.
Source