Network biology to artificial intelligence: building the next generation of predictive drug discovery

Abstract

Drug discovery remains constrained by high attrition rates, prolonged timelines, and limited ability to translate molecular insights into effective therapies. Although reductionist approaches targeting individual molecules have delivered transformative medicines, successful patient-centric treatment remains a major challenge. Furthermore, increasing evidence demonstrates that therapeutic outcomes emerge from complex, dynamic, and context-dependent biological networks. Systems biology has provided a framework to understand these interactions, yet its predictive capability has been limited by biological complexity, incomplete mechanistic knowledge, and the predominance of associative omics data. Recent advances in artificial intelligence (AI), including machine learning, deep learning, and multimodal foundation models, now offer unprecedented opportunities to integrate increasingly large and heterogeneous biomedical datasets and uncover complex biological relationships. The convergence of AI with systems biology may provide a framework for developing mechanistically informed and testable predictive models that extend beyond target identification toward understanding disease mechanisms, therapeutic responses, and treatment failures. However, mechanistic integration should not be assumed to improve prediction universally. Its value must be evaluated against appropriately matched data-driven models under defined biological conditions. Ultimately, AI-driven systems biology approaches could contribute to the development of dynamic, patient-specific computational representations of biological systems, potentially accelerating precision medicine and transforming drug discovery from empirical experimentation toward predictive, mechanism-guided therapeutic design.