Large Language Models for Zero-shot Inference of Causal Structures in Biology

2025-03-06

AI Summary

This work addresses the zero-shot causal structure inference problem among biological entities (e.g., genes, proteins), where no prior domain-specific training data is available. We propose the first evaluation paradigm that directly compares large language model (LLM) outputs against ground-truth interventional experimental data. Our method integrates customized multi-strategy prompting, retrieval-augmented generation (RAG), and lightweight LLM fine-tuning, and is systematically validated across >100 biological variables and >1,000 causal hypotheses. Results demonstrate that the model significantly outperforms random baselines, accurately capturing cell-context-specific and mechanistically complex causal relationships. Notably, it achieves effective knowledge distillation and causal hypothesis orchestration without domain-specific fine-tuning. This study provides the first empirical evidence that LLMs can reconstruct biological causal networks in a fully unsupervised setting, establishing a novel paradigm for AI-driven mechanistic discovery.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Genes, proteins and other biological entities influence one another via causal molecular networks. Causal relationships in such networks are mediated by complex and diverse mechanisms, through latent variables, and are often specific to cellular context. It remains challenging to characterise such networks in practice. Here, we present a novel framework to evaluate large language models (LLMs) for zero-shot inference of causal relationships in biology. In particular, we systematically evaluate causal claims obtained from an LLM using real-world interventional data. This is done over one hundred variables and thousands of causal hypotheses. Furthermore, we consider several prompting and retrieval-augmentation strategies, including large, and potentially conflicting, collections of scientific articles. Our results show that with tailored augmentation and prompting, even relatively small LLMs can capture meaningful aspects of causal structure in biological systems. This supports the notion that LLMs could act as orchestration tools in biological discovery, by helping to distil current knowledge in ways amenable to downstream analysis. Our approach to assessing LLMs with respect to experimental data is relevant for a broad range of problems at the intersection of causal learning, LLMs and scientific discovery.

Problem

• Infer causal relationships in biological molecular networks.
• Evaluate LLMs using real-world interventional biological data.
• Assess LLMs' ability to distill knowledge for biological discovery.

Innovation

• Uses large language models for zero-shot inference
• Evaluates causal claims with real-world interventional data
• Implements tailored prompting and retrieval-augmentation strategies