Key points are not available for this paper at this time.
Abstract Purpose of Review The ability to predict drug efficacy and toxicity holds transformative potential for drug development, offering the promise of reducing late-stage failures and accelerating the delivery of effective therapies to patients in need. Yet, despite substantial technical progress, the full potential of predictive modeling has not been realized. Developing credible and actionable predictive models remains a deeply challenging endeavor. This review presents several practical perspectives aimed at enhancing the power and reliability of predictive modeling in biomedical research. Recent Findings Success in predictive modeling hinges on a strong foundation in traditional disciplines such as physiology, pharmacology, and molecular biology, coupled with the strategic application of modern computational tools, including Quantitative Systems Pharmacology (QSP), machine learning (ML), and systems biology. The rigorous integration of experimental data and computational modeling has been increasingly recognized as essential. Moreover, effective multidisciplinary collaboration is now widely regarded as a cornerstone for building credible and impactful models. In parallel, the scientific community is actively investigating strategies to strengthen both the credibility and predictive performance of biomedical models. Summary This review outlines several practical perspectives on enhancing predictive modeling, including the capture of emergent behaviors across biological scales, the integration of foundational biomedical knowledge, the synthesis of diverse theoretical approaches, and the blending of quantitative rigor with qualitative system features. We further emphasize the value of thoughtfully reusing and adapting existing models and the importance of setting realistic expectations about model capabilities. Finally, we highlight the critical role of sustained, community-driven efforts to improve model transparency, reproducibility, and trustworthiness.
Zhang et al. (Sat,) studied this question.