Exploring MD-LOVIs, DIVATI and DRAGON descriptors for thermolysin inhibition modeling

Read full text Summarise with VIRAT
Authors: Yudith Cañizares-Carmenate, Roberto Díaz-Amador, Juan A. Castillo‐Garit
Cited by: 0
References: 39

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Currently, thousands of molecular descriptors and various software for their calculation have been defined. However, their utility is often associated with the biological problem being modeled. In this study, mathematical models of multiple linear regression are developed and validated using QSARIns software. These models correlate the chemical structure of a series of compounds with the inhibitory activity of thermolysin (expressed as pKi), serving as a model for human vasopeptidases that regulate blood pressure. Models were proposed by combining the molecular descriptors DIVATI and MD_Lovis for the multivariate analysis of thermolysin inhibition, along with a genetic algorithm for selecting an optimal set of variables. The results obtained were inferior to those from DRAGON (previously reported), however by combining the descriptors from the three models, a better fit (R²=0.7690), robustness (R²-Q² LOO =0.0318), stability (Q² LOO =0.7372 and Q² LMO =0.7300), and predictive power (R² PS =0.7612) were achieved. The application domain was determined by confirming that the predictions lie within the experimental chemical space of the database and are reliable. Additionally, the randomization test of the response variable demonstrated that the correlations are not a product of chance correlation. Considering the results obtained, it can be concluded that using different families of molecular descriptors is not mutually exclusive. In this regard, the best combinations that describe the modeled biological problem should be explored.

Fields of Study & Subjects

Computational Drug Discovery Methods Graph theory and applications Machine Learning in Materials Science

Publication details

Journal / source
Letters in Drug Design & Discovery
Volume
23
Issue
2
Pages
100380–100380
Publication date
1 February 2026
Language
EN