PerturbPFN: Probing the Limits of Synthetic Priors in Drug Perturbation Modelling

2026-07-26

AI Summary

Predicting cellular responses to unseen chemical perturbations is challenging due to unknown targets and mechanisms, high-dimensional gene expression profiles, and scarce experimental data. This work proposes PerturbPFN, the first framework to integrate synthetic structural priors with structural causal models (SCMs) within a Prior-Function Network (PFN) architecture for context-aware learning. By inferring latent regulatory graphs, sparse intervention targets, and their strengths—and propagating perturbation effects through an SCM-based decoder—the method enables efficient prediction without requiring gradient updates at test time. Combining graph neural networks with a biologically inspired synthetic data simulator, PerturbPFN achieves competitive predictive performance on both real single-cell and synthetic benchmarks while accurately recovering intervention targets, strengths, and regulatory structures, thereby balancing interpretability and computational efficiency.

Abstract

Predicting cellular responses to unseen chemical perturbations is challenging due to unknown targets and mechanisms, high-dimensional expression responses, and limited experimental coverage of the large small-molecule design space. We propose PerturbPFN, a PFN-style amortized model for unknown-target perturbation prediction under a hierarchical synthetic structural prior. Instead of directly regressing high-dimensional expression responses, PerturbPFN infers a latent system graph, sparse atomic intervention targets, and intervention strengths, then propagates their effects through an SCM decoder. The model is trained entirely on prior-predictive synthetic episodes generated from biologically motivated graph and expression simulators, enabling structured in-context learning without test-time gradient updates. We evaluate PerturbPFN on both real single-cell perturbation data and synthetic benchmarks, covering effect prediction, target identification, and regulatory structure discovery. Our results show that PerturbPFN offers a complementary trade-off to specialized baselines, achieving competitive perturbation prediction with low inference cost while exposing interpretable intermediate estimates of targets, strengths, and system structure.

Problem

• drug perturbation
• cellular response prediction
• unknown targets
• high-dimensional expression
• small-molecule design space

Innovation

• PerturbPFN
• synthetic prior
• structural causal model
• amortized inference
• perturbation prediction