{"entity":{"id":"gears","kind":"technology","name":"GEARS and perturbation prediction benchmarks","aka":[],"tldr":"GEARS is a graph model predicting the effect of gene knockouts; the perturbation benchmarks around it showed how hard the problem is.","summary":"GEARS is a graph neural network from Stanford that predicts the transcriptional effect of gene perturbations by reasoning over gene-gene relationship graphs, which lets it extrapolate to perturbations, including combinations, it has never seen. The Nature Biotechnology 2023 paper introduced the approach and set an early standard for the task. Its lasting influence is on evaluation: benchmark studies in 2024 and 2025 found that many single-cell foundation models barely beat simple baselines such as predicting the mean response, which exposed weak baselines across the field and sharpened standards through efforts like the Virtual Cell Challenge. Combinatorial perturbation prediction remains its distinctive strength. For a newcomer: GEARS predicts what happens when you knock out a gene, and the contests built around it showed how hard that prediction really is.","status":"emerging","asOf":"2026-09-08","links":[{"label":"Nature Biotechnology 2023","url":"https://doi.org/10.1038/s41587-023-01905-6"}],"tags":["virtual-cell","benchmark"],"related":[],"cancers":[],"sections":["ai-computation"],"technologies":["crispr-screens"],"targets":[],"drugs":[],"companies":[],"institutions":["stanford"],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":["paper-roohani-nat-biotechnol"],"journals":[],"dependsOn":[],"notes":[],"principle":"GEARS is a graph neural network over gene relationships.","strengths":["Combinatorial perturbations"],"limitations":["Weak baselines exposed field-wide"],"since":2023},"route":"/technologies/gears/","neighbours":{"section":[{"id":"ai-computation","kind":"section","name":"AI & Computation","route":"/fronts/ai-computation/"}],"technology":[{"id":"crispr-screens","kind":"technology","name":"CRISPR functional genomics","route":"/technologies/crispr-screens/"}],"institution":[{"id":"stanford","kind":"institution","name":"Stanford Health Care / Stanford Cancer Institute","route":"/institutions/stanford/"}],"paper":[{"id":"paper-roohani-nat-biotechnol","kind":"paper","name":"Predicting transcriptional outcomes of novel multigene perturbations with GEARS","route":"/key-papers/paper-roohani-nat-biotechnol/"}],"roadmap":[{"id":"virtual-cell","kind":"roadmap","name":"Virtual cell roadmap: from bulk omics to a predictive model of a cancer cell","route":"/roadmaps/virtual-cell/"}]}}