PertMind: Eliciting Emergent Biological Reasoning in LLM via Reinforcement Learning on Cellular Perturbation Data

2026-08-17

AI Summary

This study addresses the bottleneck of expensive manual annotation in post-training large models for biological reasoning by proposing a novel paradigm that leverages cellular perturbation atlases to construct reinforcement learning environments. Through credible trajectory-supervised initialization, gene-pathway multi-scale rewards, and forward perturbation prediction, we achieve autonomous training driven entirely by experimental data. This approach generates high-quality multi-scale biological representations without additional fine-tuning, significantly enhancing reasoning capabilities in unseen scenarios. Furthermore, it demonstrates successful zero-shot transferability across diverse downstream tasks, validating both the feasibility and scalability of utilizing large-scale experimental data to drive general-purpose biological reasoning.

Abstract

Large language models can describe mechanisms, yet scalable post-training still depends on costly, manually curated biological reasoning traces. Here we show that cellular perturbation atlases can instead become reinforcement-learning environments, where measured gene responses provide computable rewards for biological reasoning. We introduce PertMind, which combines trusted-trajectory supervised initialization with gene-, pathway-, and format-level reinforcement signals. Trained only on forward perturbation-response prediction, PertMind improved response inference in unseen cellular contexts while retaining general language capabilities. It also transferred without task-specific post-training to reverse perturbation identification, double-perturbation reasoning, phenotypic-screen prioritization, and biological-process interpretation. PertMind further generated biological profiles that supported competitive gene, cell, and donor representations across multiscale downstream tasks. These results support the hypothesis that reinforcement on experimental endpoints can concentrate reusable biological strategies already accessible to pretrained models. More broadly, perturbation-derived reinforcement learning offers a scalable route for transforming expanding experimental atlases into training environments for general-purpose biological reasoning.

Problem

• Biological Reasoning
• Large Language Models
• Post-training Scalability
• Cellular Perturbation

Innovation

• Reinforcement Learning
• Cellular Perturbation
• Biological Reasoning
• PertMind
• Zero-shot Transfer