LLM-Guided Retrieval for Prediction of Molecular Perturbation Responses

2026-08-03

AI Summary

Predicting transcriptional responses to small-molecule perturbations across diverse cell lines is hindered by limited experimental coverage. This work proposes a novel retrieval-aggregation paradigm that leverages large language models (LLMs) as constrained biological priors to retrieve compounds with semantically relevant biological profiles in the target cell line and aggregates their expression profiles via averaging to predict responses to unmeasured drugs. Notably, this approach is the first to employ LLMs for prior-guided retrieval in perturbation prediction, emphasizing that high-quality retrieval outperforms complex predictors and substantially enhances zero-shot generalization. The method consistently surpasses existing baselines under unseen-drug, unseen-cell-line, and open-world settings, demonstrating particularly strong performance in cross-cell-line scenarios with higher correlation, lower error, and more accurate prediction of gene regulatory directions.

Abstract

Predicting transcriptomic responses to small-molecule perturbations across cell lines is central to drug discovery, but exhaustive profiling of drug-cell combinations is infeasible. We frame molecular perturbation prediction as retrieve-and-aggregate: approximate an unmeasured drug's response in a cell line by aggregating measured responses of a small set of biologically related compounds. We propose LLM-Guided Retrieval (LGR), where a large language model (LLM) ranks candidate neighbor drugs (restricted to those profiled in the target cell line); after which a fixed mean aggregator combines their observed expression deltas to form the prediction. We evaluate on the Tahoe-100M single-cell perturbation atlas under unseen-drug, unseen-cell-line, and open-world regimes. LGR consistently improves over drug mean, ChemCPA, and chemistry-based kNN baselines, with the strongest gains for unseen cell-line generalization, where it achieves higher correlation and lower error than mean baselines. Across settings, LGR improves directional (sign) accuracy of gene regulation, indicating better recovery of biologically meaningful perturbation effects even when magnitude-based metrics are similar. These results suggest that retrieval quality, rather than predictor complexity, is a key driver of zero-shot molecular perturbation prediction, and that LLMs can provide a useful biological prior when used as constrained retrieval modules.

Problem

• molecular perturbation prediction
• transcriptomic response
• drug-cell combinations
• zero-shot prediction
• single-cell perturbation

Innovation

• LLM-Guided Retrieval
• molecular perturbation prediction
• zero-shot generalization
• biological prior
• retrieve-and-aggregate