New method for systems pharmacology modeling and its application for assessing the effectiveness of diabetes treatment

Mathematical modeling has been an integral part of the drug development process for several decades. The growing complexity of challenges facing modern medicine, combined with the rapid advancement of technology, has led to the emergence of new modeling methods, one of which is quantitative systems pharmacology (QSP) modeling. QSP modeling is used to describe the dynamical response of a biological system to the administration of existing or theoretical drugs. QSP models typically consist of ordinary differential equations that describe the relationship between drug concentration (pharmacokinetics) and the concentration of various biomarkers (pharmacodynamics) in organs and tissues. The potential spectrum of applications of such models in drug development is broad; however, their practical application at the level of regulatory agencies is complicated by the lack of standards for developing QSP models, specifically: the absence of a systematic approach and reproducibility in data collection and analysis necessary for the development of mechanistic models, the lack of unified criteria for assessing model quality, and the limited ability to account for real variability in clinical responses when solving the direct problem. Thus, the aim of this work was to develop a method for systems pharmacology modeling, implement it in software, and apply it to address drug development questions in the treatment of diabetes. In this research, a systematization of heuristic and quantitative rules and criteria, as well as numerical and statistical methods for data analysis, was conducted to formulate a unified approach to the development of QSP models with the goal of improving their quality, reproducibility, and development speed. The proposed method includes a model-oriented approach to data search and analysis, compilation of datasets sufficient for solving the inverse problem and selecting optimal model structure, model validation, and a new method for predicting the clinical efficacy and safety of drugs. Based on the proposed method, two mechanistic models of diabetes were developed. The renal glucose reabsorption model provided an explanation for experimentally observed contradictions in the amount of glucose excreted under the influence of dapagliflozin, canagliflozin, and empagliflozin in patients with type 2 diabetes. Using the second integrative model of glucose homeostasis, individual patient characteristics influencing the efficacy of gliflozins were identified. Finally, the proposed method for developing QSP models served as the basis for the software "Simurg" (Registered with the Ministry of Digital Development, Communications, and Mass Media of the Russian Federation, No. 22787) for addressing quantitative pharmacology tasks and developing new mathematical models.