Agents based on large language models have shown great potential in accelerating scientific discovery by leveraging their rich background knowledge and reasoning capabilities. In this paper, we introduce BioDiscoveryAgent, an agent that designs new experiments, reasons about their outcomes, and efficiently navigates the hypothesis space to reach desired solutions. We demonstrate our agent on the problem of designing genetic perturbation experiments, where the aim is to find a small subset out of many possible genes that, when perturbed, result in a specific phenotype (e.g., cell growth). Utilizing its biological knowledge, BioDiscoveryAgent can uniquely design new experiments without the need to train a machine learning model or explicitly design an acquisition function as in Bayesian optimization. Moreover, BioDiscoveryAgent using Claude 3.5 Sonnet achieves an average of 21% improvement in predicting relevant genetic perturbations across six datasets, and a 46% improvement in the harder task of non-essential gene perturbation, compared to existing Bayesian optimization baselines specifically trained for this task. Our evaluation includes one dataset that is unpublished, ensuring it is not part of the language model's training data. Additionally, BioDiscoveryAgent predicts gene combinations to perturb more than twice as accurately as a random baseline, a task so far not explored in the context of closed-loop experiment design. The agent also has access to tools for searching the biomedical literature, executing code to analyze biological datasets, and prompting another agent to critically evaluate its predictions. Overall, BioDiscoveryAgent is interpretable at every stage, representing an accessible new paradigm in the computational design of biological experiments with the potential to augment scientists' efficacy.

SUMMARY: The Probabilistic Identification of Causal SNPs (PICS) algorithm and web application was developed as a fine-mapping tool to determine the likelihood that each single nucleotide polymorphism (SNP) in LD with a reported index SNP is a true causal polymorphism. PICS is notable for its ability to identify candidate causal SNPs within a locus using only the index SNP, which are widely available from published GWAS, whereas other methods require full summary statistics or full genotype data. However, the original PICS web application operates on a single SNP at a time, with slow performance, severely limiting its usability. We have developed a next-generation PICS tool, PICS2, which enables performance of PICS analyses of large batches of index SNPs with much faster performance. Additional updates and extensions include use of LD reference data generated from 1000 Genomes phase 3; annotation of variant consequences; annotation of GTEx eQTL genes and downloadable PICS SNPs from GTEx eQTLs; the option of generating PICS probabilities from experimental summary statistics; and generation of PICS SNPs from all SNPs of the GWAS catalog, automatically updated weekly. These free and easy-to-use resources will enable efficient determination of candidate loci for biological studies to investigate the true causal variants underlying disease processes. AVAILABILITY AND IMPLEMENTATION: PICS2 is available at https://pics2.ucsf.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.