SynthPert: Enhancing LLM Biological Reasoning via Synthetic Reasoning Traces for Cellular Perturbation Prediction

2025-09-29

AI Summary

Predicting cellular responses to genetic perturbations remains a central challenge in systems biology and therapeutic discovery. While large language models (LLMs) show emerging promise for biological reasoning, their adaptation to structured perturbation experimental data is limited. To address this, we propose a supervised fine-tuning framework grounded in synthetic reasoning trajectories: state-of-the-art models generate biologically informed reasoning chains, which are rigorously quality-filtered and used to distill implicit domain knowledge—thereby substantially enhancing LLMs’ generalization capability for perturbation response modeling. Evaluated on the PerturbQA benchmark, our method achieves state-of-the-art performance using only 2% high-quality training data and attains 87% accuracy on unseen RPE1 cell types, effectively overcoming key bottlenecks in data efficiency and cross-cell-type generalization.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Predicting cellular responses to genetic perturbations represents a fundamental challenge in systems biology, critical for advancing therapeutic discovery and virtual cell modeling. While large language models (LLMs) show promise for biological reasoning, their application to perturbation prediction remains underexplored due to challenges in adapting them to structured experimental data. We present SynthPert, a novel method that enhances LLM performance through supervised fine-tuning on synthetic reasoning traces generated by frontier models. Using the PerturbQA benchmark, we demonstrate that our approach not only achieves state-of-the-art performance but surpasses the capabilities of the frontier model that generated the training data. Our results reveal three key insights: (1) Synthetic reasoning traces effectively distill biological knowledge even when partially inaccurate, (2) This approach enables cross-cell-type generalization with 87% accuracy on unseen RPE1 cells, and (3) Performance gains persist despite using only 2% of quality-filtered training data. This work shows the effectiveness of synthetic reasoning distillation for enhancing domain-specific reasoning in LLMs.

Problem

• Predicting cellular responses to genetic perturbations in systems biology
• Enhancing LLM biological reasoning with synthetic reasoning traces
• Overcoming challenges in adapting LLMs to structured experimental data

Innovation

• Synthetic reasoning traces enhance LLM biological reasoning
• Supervised fine-tuning improves cellular perturbation prediction
• Method achieves generalization with minimal quality-filtered data