Arc Institute

Overview

Arc Institute is a nonprofit research organization pursuing foundational questions at the intersection of biology, computation, and AI. It runs the Virtual Cell Initiative and the Alzheimer's Disease Initiative. Its Computational Technology Center counts Hani Goodarzi as a founding member; Goodarzi also holds faculty at UCSF. Arc maintains an on-premises GPU cluster and open-sources all of its research.

Key models from Arc include EVO 2, an open-source genomics model trained on the human reference genome, and the STACK architecture, which combines observational and causal biological data. Arc also published SC Base Count, an AI-agent reprocessing of the NIH Sequence Read Archive. Yusuf Roohani (Assistant Director of ML) authored the STACK paper and led the SC Base Count project. Roohani's perspective is that AlphaFold-style scaling will not generalize to most biological problems; Arc's approach uses lab-in-the-loop active learning to generate higher-quality causal data rather than relying solely on scale.

Sign in to read the full article.

Sign In