Xingran Zhao , Yanbu Guo , Bingyi Wang , Weihua Li
{"title":"CKDTA:用于药物靶点亲和力预测的化学知识增强框架","authors":"Xingran Zhao , Yanbu Guo , Bingyi Wang , Weihua Li","doi":"10.1016/j.jocs.2025.102706","DOIUrl":null,"url":null,"abstract":"<div><div>Accurate drug–target affinity (DTA) prediction is a cornerstone of efficient drug discovery, as it directly accelerates the screening of potential therapeutic candidates, reduces the cost of preclinical experiments, and shortens the development cycle of new drugs. However, existing deep learning-based methods face two main challenges: (I) Purely data-driven approaches struggle to capture the functional semantics of molecules, such as the role of specific functional regions and chemical element properties in binding interactions, due to the lack of integration with chemical prior knowledge, leading to unreliable predictions; (II) the integration of topological structure from graphs and long-range dependencies from sequences is insufficient, often failing to capture complementary features, limiting the model’s generalization ability, especially for novel drugs or targets commonly encountered in early drug discovery . To address these issues, we propose <strong>CKDTA</strong>, a <strong>C</strong>hemical <strong>K</strong>nowledge Enhanced framework for <strong>D</strong>rug-<strong>T</strong>arget <strong>A</strong>ffinity prediction. Our framework introduces two key innovations: (1) a chemical knowledge-enhanced molecular modeling approach, which constructs a multi-layer molecular graph incorporating atom-level features, chemical element information, and functional regions, enabling the capture of functional semantics through a hierarchical attention mechanism, while leveraging chemical prior knowledge; (2) a co-attention module designed to optimize sequence interaction information by leveraging graph-based interaction data, compensating for the lack of spatial structural information in sequence data. This module fully exploits the topological structure of graphs and the long-range dependencies in sequences, capturing complementary features. Extensive experiments on benchmark datasets demonstrate that CKDTA outperforms state-of-the-art methods. Furthermore, cold-start experiments validate its generalizability, highlighting its potential for drug discovery applications.</div></div>","PeriodicalId":48907,"journal":{"name":"Journal of Computational Science","volume":"92 ","pages":"Article 102706"},"PeriodicalIF":3.7000,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"CKDTA: A chemical knowledge-enhanced framework for drug–target affinity prediction\",\"authors\":\"Xingran Zhao , Yanbu Guo , Bingyi Wang , Weihua Li\",\"doi\":\"10.1016/j.jocs.2025.102706\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Accurate drug–target affinity (DTA) prediction is a cornerstone of efficient drug discovery, as it directly accelerates the screening of potential therapeutic candidates, reduces the cost of preclinical experiments, and shortens the development cycle of new drugs. However, existing deep learning-based methods face two main challenges: (I) Purely data-driven approaches struggle to capture the functional semantics of molecules, such as the role of specific functional regions and chemical element properties in binding interactions, due to the lack of integration with chemical prior knowledge, leading to unreliable predictions; (II) the integration of topological structure from graphs and long-range dependencies from sequences is insufficient, often failing to capture complementary features, limiting the model’s generalization ability, especially for novel drugs or targets commonly encountered in early drug discovery . To address these issues, we propose <strong>CKDTA</strong>, a <strong>C</strong>hemical <strong>K</strong>nowledge Enhanced framework for <strong>D</strong>rug-<strong>T</strong>arget <strong>A</strong>ffinity prediction. Our framework introduces two key innovations: (1) a chemical knowledge-enhanced molecular modeling approach, which constructs a multi-layer molecular graph incorporating atom-level features, chemical element information, and functional regions, enabling the capture of functional semantics through a hierarchical attention mechanism, while leveraging chemical prior knowledge; (2) a co-attention module designed to optimize sequence interaction information by leveraging graph-based interaction data, compensating for the lack of spatial structural information in sequence data. This module fully exploits the topological structure of graphs and the long-range dependencies in sequences, capturing complementary features. Extensive experiments on benchmark datasets demonstrate that CKDTA outperforms state-of-the-art methods. Furthermore, cold-start experiments validate its generalizability, highlighting its potential for drug discovery applications.</div></div>\",\"PeriodicalId\":48907,\"journal\":{\"name\":\"Journal of Computational Science\",\"volume\":\"92 \",\"pages\":\"Article 102706\"},\"PeriodicalIF\":3.7000,\"publicationDate\":\"2025-09-15\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Computational Science\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S1877750325001838\",\"RegionNum\":3,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Computational Science","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1877750325001838","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS","Score":null,"Total":0}
CKDTA: A chemical knowledge-enhanced framework for drug–target affinity prediction
Accurate drug–target affinity (DTA) prediction is a cornerstone of efficient drug discovery, as it directly accelerates the screening of potential therapeutic candidates, reduces the cost of preclinical experiments, and shortens the development cycle of new drugs. However, existing deep learning-based methods face two main challenges: (I) Purely data-driven approaches struggle to capture the functional semantics of molecules, such as the role of specific functional regions and chemical element properties in binding interactions, due to the lack of integration with chemical prior knowledge, leading to unreliable predictions; (II) the integration of topological structure from graphs and long-range dependencies from sequences is insufficient, often failing to capture complementary features, limiting the model’s generalization ability, especially for novel drugs or targets commonly encountered in early drug discovery . To address these issues, we propose CKDTA, a Chemical Knowledge Enhanced framework for Drug-Target Affinity prediction. Our framework introduces two key innovations: (1) a chemical knowledge-enhanced molecular modeling approach, which constructs a multi-layer molecular graph incorporating atom-level features, chemical element information, and functional regions, enabling the capture of functional semantics through a hierarchical attention mechanism, while leveraging chemical prior knowledge; (2) a co-attention module designed to optimize sequence interaction information by leveraging graph-based interaction data, compensating for the lack of spatial structural information in sequence data. This module fully exploits the topological structure of graphs and the long-range dependencies in sequences, capturing complementary features. Extensive experiments on benchmark datasets demonstrate that CKDTA outperforms state-of-the-art methods. Furthermore, cold-start experiments validate its generalizability, highlighting its potential for drug discovery applications.
期刊介绍:
Computational Science is a rapidly growing multi- and interdisciplinary field that uses advanced computing and data analysis to understand and solve complex problems. It has reached a level of predictive capability that now firmly complements the traditional pillars of experimentation and theory.
The recent advances in experimental techniques such as detectors, on-line sensor networks and high-resolution imaging techniques, have opened up new windows into physical and biological processes at many levels of detail. The resulting data explosion allows for detailed data driven modeling and simulation.
This new discipline in science combines computational thinking, modern computational methods, devices and collateral technologies to address problems far beyond the scope of traditional numerical methods.
Computational science typically unifies three distinct elements:
• Modeling, Algorithms and Simulations (e.g. numerical and non-numerical, discrete and continuous);
• Software developed to solve science (e.g., biological, physical, and social), engineering, medicine, and humanities problems;
• Computer and information science that develops and optimizes the advanced system hardware, software, networking, and data management components (e.g. problem solving environments).