跳转至

药物毒性预测中提示词工程的分析研究

文章背景与核心概要

临床试验成本极其高昂（高达 130 万英镑），且失败率高达 90% 左右，其中药物毒性是导致失败的主要原因。尽管大型语言模型（LLM）被越来越多地用于预测药物毒性，但其输出结果对提示词措辞的微小变化往往高度敏感。

本文深入探讨了提示词工程对药物毒性预测的真实影响。研究人员测试了在各种提示词结构（评估工作角色、提示词结构以及规则解释）下 LLM 的表现，以识别用于预测的重要化学性质。研究结果表明，LLM 固有的自然方差超出了提示词微调所带来的收益。然而，使用传统的化学信息学代码来提取特征（而非依赖 LLM 生成的数值），能够显著提升下游机器学习模型的性能。

摘要 (Summary)

Clinical trials are exceptionally costly—reaching up to £1.3 million—with a high failure rate of approximately 90%, largely driven by drug toxicity. While Large Language Models (LLMs) are increasingly used to predict toxicity, their outputs can be highly sensitive to minor changes in prompt phrasing.

This paper investigates the true impact of prompt engineering on drug toxicity prediction. The researchers tested LLMs under various prompt constructs—evaluating job roles, prompt structures, and rule interpretations—to identify significant chemical properties for prediction. The findings reveal that the natural variance inherent in LLMs outweighs the benefits of prompt fine-tuning. However, utilizing traditional chemoinformatic code to extract features, rather than relying on LLM-generated values, resulted in substantial performance improvements for downstream machine learning models.

临床试验成本极其高昂——高达 130 万英镑——失败率约为 90%，这在很大程度上是由药物毒性引起的。虽然大语言模型（LLM）被越来越多地用于预测毒性，但它们的输出对提示词措辞的微小变化非常敏感。

本文研究了提示词工程对药物毒性预测的真实影响。研究人员测试了不同提示词结构下的 LLM——评估了工作角色、提示词结构和规则解释——以识别预测中具有重要意义的化学性质。研究结果表明，LLM 固有的自然方差大于提示词微调带来的好处。然而，利用传统的化学信息学代码来提取特征，而不是依赖 LLM 生成的数值，显著提升了下游机器学习模型的性能。

论文元数据 (Paper Metadata)

  • arXiv ID: arXiv:2609.03635 [cs.AI]
  • Subject: Artificial Intelligence (cs.AI)
  • Authors: Mia MacGregor, Aakash Welgamage Don, Mark Bartlett
  • Submission Date: September 3, 2026
  • Conference Acceptance: Accepted at the CIBB 2026 conference (Website)
  • DOI: 10.48550/arXiv.2609.03635
  • arXiv ID: arXiv:2609.03635 [cs.AI]
  • 学科领域: 人工智能 (cs.AI)
  • 作者: Mia MacGregor, Aakash Welgamage Don, Mark Bartlett
  • 提交日期: 2026年9月3日
  • 会议录用: 已被 CIBB 2026 会议接收 (网站)
  • DOI: 10.48550/arXiv.2609.03635

摘要正文 (Abstract)

Clinical trials in the UK can cost up to £1.3 million, with approximately 90% drug failure rate. Toxicity is a major contributing factor in drug failure. Testing is time and cost intensive. In recent years, the use of artificial intelligence has been increasingly explored to aid in the prediction of drug toxicity, with extensive use of large language models (LLMs). However, LLMs can show considerable variation when minor changes are made to prompts, which raises concerns about their sensitivity to prompt engineering. Prompt engineering is used to optimise a prompt given to an LLM to generate the desired output.

This paper proposes a method to analyse prompt engineering for drug toxicity prediction. The aim of the paper is to investigate the importance of prompt phrasing for drug toxicity prediction. LLMs were prompted to identify chemical properties of significance when predicting drug toxicity. Prompts were constructed to investigate: job role, prompt structuring, and rule interpretation. LLMs were then used to generate datasets, using the identified features from initial prompting, which were then passed to machine learning algorithms.

The experiments show that the natural variance which occurs in LLMs outweighs any fine-tuning of prompts. There were, however, substantial improvements in model performance when using chemoinformatic code to extract features instead of using LLM-generated values. The proposed analysis methodology is applicable to a wide range of prompt types across different areas of bioinformatics.

英国的临床试验成本高达 130 万英镑，药物失败率约为 90%。毒性是导致药物失败的主要因素。测试过程既耗时又耗资。近年来，人们越来越多地探索使用人工智能来辅助预测药物毒性，并广泛使用了大语言模型（LLM）。然而，当提示词发生微小变化时，LLM 可能会表现出相当大的差异，这引发了人们对其对提示词工程敏感性的担忧。提示词工程用于优化赋予 LLM 的提示词以生成所需的输出。

本文提出了一种分析用于药物毒性预测的提示词工程的方法。本文旨在研究提示词措辞对于药物毒性预测的重要性。通过提示 LLM，使其在预测药物毒性时识别具有显著意义的化学性质。构建提示词旨在研究：工作角色、提示词结构和规则解释。随后，LLM 被用于利用从初始提示词中识别出的特征来生成数据集，并将其传递给机器学习算法。

实验表明，LLM 中发生的自然方差超过了对提示词的任何微调。然而，当使用化学信息学代码来提取特征而不是使用 LLM 生成的数值时，模型的性能有了显著提升。所提出的分析方法适用于生物信息学不同领域中的广泛提示词类型。

访问与资源 (Access & Resources)