ScreenShot: A Foundation Model for Few-Shot Combination Drug Screening

2026-08-12

AI Summary

Combinatorial drug screening is hindered by an expansive search space, high costs, and scarce patient samples, while existing approaches rely heavily on molecular profiling and cohort-specific training, limiting their clinical applicability. This work proposes ScreenShot, a hierarchical Transformer-based foundation model pretrained on 40 drug screening datasets, which leverages in-context learning to predict individual patient responses to combination therapies directly from minimal functional assay data—without requiring molecular profiles or task-specific fine-tuning. ScreenShot achieves, for the first time, few-shot prediction of combinatorial drug responses in a zero-fine-tuning, profile-free setting, and integrates a weighted k-means++ active learning strategy to guide experimental design. Evaluated on four independent test sets, ScreenShot outperforms baseline methods in both response prediction accuracy and identification of effective treatments; its active learning strategy attains the hit-detection performance of uniform screening using only one-third of the experimental budget.

Abstract

Treating patients with combinations of drugs reduces the risk of resistance to any individual drug. Finding effective combinations is difficult because the large search space makes combinatorial screens prohibitively expensive, time consuming, and often technically infeasible. Predictive models can fill this gap, yet existing methods typically require molecular profiling of each sample and per-cohort training, limiting their applicability when time and tissue are scarce. To address this challenge, we introduce ScreenShot, a hierarchical transformer pretrained on 40 drug screening datasets covering 3,700 drugs and 6,000 biological samples, whose architecture mirrors the nested structure of screening data. Given a few-shot context of observations from a new patient, ScreenShot predicts the response of the sample to combination therapies through in-context learning, operating directly on functional measurements with no fine-tuning and no molecular profiling. On four held-out datasets, ScreenShot outperforms all baselines in both prediction accuracy and identification of selectively effective treatments. ScreenShot's internal representations are directly useful for experimental design: we use them to drive a weighted k-means++ active learning strategy that selects which experiments to run, achieving the same hit detection as uniform screening with a third of the budget. Source code and interactive dashboard: https://github.com/tansey-lab/screenshot.

Problem

• combination drug screening
• few-shot learning
• predictive modeling
• drug resistance
• high-throughput screening

Innovation

• foundation model
• few-shot learning
• combination drug screening
• in-context learning
• active learning