《生命科学》 2025, 37(12): 1563-1572

人工智能大语言模型在药物靶点发现中的应用

冯　毅 , 马小洁 , 吴艳玲^* , 应天雷^*

复旦大学基础医学院，上海 200032

摘　要：

药物靶点发现是现代药物研发的核心环节，然而，传统的生化筛选、组学分析等方法因难度大、成本高而应用受限。随着人工智能大语言模型的快速发展，药物靶点发现迎来了新的机遇。在药物靶点挖掘过程中，学习人类语言的自然语言模型能够高效整合和全面分析文献资料，识别与疾病相关的关键生物学途径及靶点。此外，通过对生物“语言”的学习，基因组学大语言模型提升了对致病变异和基因表达的预测能力；转录组学大语言模型可系统构建基因调控网络；蛋白质组学大语言模型在蛋白质结构、功能及互作预测中展现出巨大潜力；单细胞多组学大语言模型整合不同组学技术信息。这些大语言模型为药物靶点发现提供丰富的生物学信息，加速发现具有强大潜力的候选药物靶点。本文综述了大语言模型在药物靶点发现中的最新应用，并深入探讨其面临的挑战及未来发展方向。

通讯作者：吴艳玲 , Email:yanlingwu@fudan.edu.cn 应天雷 , Email:tlying@fudan.edu.cn

Application of artificial intelligence large language models in drug target discovery

FENG Yi , MA Xiao-Jie , WU Yan-Ling^* , YING Tian-Lei^*

School of Basic Medical Sciences, Fudan University, Shanghai 200032, China

Abstract:

Drug target discovery represents a fundamental and pivotal stage in modern drug development. However, traditional methods such as biochemical screening and omics analysis are limited by their high complexity and cost. With the rapid advancement of artificial intelligence large language models, new opportunities have emerged in drug target discovery. In the process of identifying drug targets, natural language models that learn human language can comprehensively analyze literature and extract key biological pathways and targets related to diseases. By learning the "language" of biology, genomics large language models have enhanced the ability to predict pathogenic variants and gene expression; transcriptomics large language models can systematically construct gene regulatory networks; proteomics large language models exhibit great potential in predicting protein structure, function, and interactions; single-cell multi-omics large language models integrate information from various omics technologies. These large language models provide abundant biological information for drug target discovery, accelerating the process of target identification and drug development. This review summarizes the application of large language models in drug target discovery and discusses the challenges in this field.

Communication Author：WU Yan-Ling , Email:yanlingwu@fudan.edu.cn YING Tian-Lei , Email:tlying@fudan.edu.cn