Molecular Modeling and Machine Learning Guided Target Fishing and Therapeutic Interventions against Cardiovascular Diseases

Abstract

Cardiovascular Disease (CVD) is a major global health issue, with thrombosis, or blood clot formation, as a key factor. A disintegrin-like and metalloproteinase with thrombospondin type 1 repeats 13 (ADAMTS13), a protease that cleaves the ultra-large von Willebrand factor (ULVWF), regulates thrombosis by preventing plug development. Despite its importance, ADAMTS13 regulation remains poorly understood. This study explores platelet plug formation, focusing on ADAMTS13's role in clot regulation. We developed a knowledge-driven biological regulatory network (BRN) and built four classification models using protein data from STRING and DISGENET to distinguish proteins linked to ADAMTS13, CVD, and thrombosis from those related only to CVD and thrombosis. The models, including support vector classifier, random forest, logistic regression, and ANN, were optimized using GridSearchCV. The logistic regression and ANN models showed strong performance, with the accuracy rates of 87.05% and 88.82%, respectively. The ANN model demonstrated a balanced performance between precision (83.82%) and recall (87.69%). Thrombin and plasmin were identified as ADAMTS13 inhibitors from BRN, offering insights into regulation and potential therapeutic targets. ADAMTS13 mRNA secondary structure was predicted using RNAfold, though reliability was limited by mRNA dynamics. This study investigates ADAMTS13 regulation and its role in thrombosis and CVD, using computational approaches to deepen understanding of the molecular mechanisms. Future work should aim to probe the regulatory mechanistic of ADAMTS13, enhance classification models performance, and predict its full-length protein structure for insights into its functional mechanism.