Machine learning approach for predicting drug-like molecules targeting Calmodulin pathway proteins

Recently, numerous models have been developed to predict drug interactions with molecules. However, integrating diverse data sources and improving the accuracy of biological activity predictions remains a challenge. This work proposes a novel solution that addresses these limitations. Here, we have...

Descripción completa

Detalles Bibliográficos
Autores: Baltasar Marchueta, Maider, López, Naia, Alicante Martínez, Sara, Barbolla Cuadrado, Iratxe, García Ibarluzea, Markel, Ramis Cortés, Rafael, Salomon, Ane Miren, Muguruza Montero, Arantza, Núñez Viadero, Eider, Leonardo Liceranzu, Aritz, Arrasate Gil, Sonia, Sotomayor Anduiza, María Nuria, Montemore, Matthew M., Villarroel, Alvaro, Bergara Jauregui, Aitor, Lete Expósito, María Esther, González Díaz, Humberto
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universidad del País Vasco
Repositorio:Addi. Archivo Digital para la Docencia y la Investigación
OAI Identifier:oai:addi.ehu.eus:10810/75600
Acceso en línea:http://hdl.handle.net/10810/75600
Access Level:acceso abierto
Palabra clave:drug discovery
Calmodulin
Riluzole
chemoinformatic
machine learning
synthesis
biological assays
docking studies
Descripción
Sumario:Recently, numerous models have been developed to predict drug interactions with molecules. However, integrating diverse data sources and improving the accuracy of biological activity predictions remains a challenge. This work proposes a novel solution that addresses these limitations. Here, we have developed a machine learning model to predict the efficacy of different assays and drugs for diseases related to calmodulin. To achieve this, we have compiled a comprehensive data set including commercialized drugs and experimental compounds targeting CaM complexes. The IFPTML-XGB model achieved high predictive performance, with a test accuracy of 89.1% and a sensitivity of 89.0%, demonstrating its robustness for assay efficacy prediction. We have used the IFPTML modeling technique to identify key factors influencing these activities. We have also synthesized novel riluzole derivatives and have tested them both experimentally and computationally. Biological assays and molecular docking studies have been performed to provide a molecular-scale picture of the molecule−CaM interaction. To validate the model’s utility, we tested it on these derivatives. We have found that the model correctly predicts which derivatives were the most bioactive, indicating that this framework can be used to identify promising candidates for new drug formulations. This research not only improves our understanding of CaM-related diseases, but also provides an effective framework for developing new treatments based on predictive modeling.