Computational Identification of Chemical Compounds with Potential Activity against Leishmania amazonensis using Nonlinear Machine Learning Techniques

Title:Computational Identification of Chemical Compounds with Potential Activity against Leishmania amazonensis using Nonlinear Machine Learning Techniques

Volume: 18 Issue: 27

Author(s): Juan Alberto Castillo-Garit*, Naivi Flores-Balmaseda, Orlando Álvarez, Hai Pham-The, Virginia Pérez-Doñate, Francisco Torrens and Facundo Pérez-Giménez

Affiliation:

• Unidad de Toxicologia Experimental, Universidad de Ciencias Medicas de Villa Clara, Santa Clara, 50200,Cuba

Keywords: Anti-leishmanial activity, Chemical compound, L. amazonensis, ML-based model, Virtual screening, WEKA.

Abstract: Leishmaniasis is a poverty-related disease endemic in 98 countries worldwide, with morbidity and mortality increasing daily. All currently used first-line and second-line drugs for the treatment of leishmaniasis exhibit several drawbacks including toxicity, high costs and route of administration. Consequently, the development of new treatments for leishmaniasis is a priority in the field of neglected tropical diseases. The aim of this work is to develop computational models those allow the identification of new chemical compounds with potential anti-leishmanial activity. A data set of 116 organic chemicals, assayed against promastigotes of Leishmania amazonensis, is used to develop the theoretical models. The cutoff value to consider a compound as active one was IC50≤1.5μM. For this study, we employed Dragon software to calculate the molecular descriptors and WEKA to obtain machine learning (ML) models. All ML models showed accuracy values between 82% and 91%, for the training set. The models developed with k-nearest neighbors and classification trees showed sensitivity values of 97% and 100%, respectively; while the models developed with artificial neural networks and support vector machine showed specificity values of 94% and 92%, respectively. In order to validate our models, an external test-set was evaluated with good behavior for all models. A virtual screening was performed and 156 compounds were identified as potential anti-leishmanial by all the ML models. This investigation highlights the merits of ML-based techniques as an alternative to other more traditional methods to find new chemical compounds with anti-leishmanial activity.

Cite this article as:

Castillo-Garit Alberto Juan *, Flores-Balmaseda Naivi , Álvarez Orlando , Pham-The Hai , Pérez-Doñate Virginia , Torrens Francisco and Pérez-Giménez Facundo , Computational Identification of Chemical Compounds with Potential Activity against Leishmania amazonensis using Nonlinear Machine Learning Techniques, Current Topics in Medicinal Chemistry 2018; 18 (27) . https://dx.doi.org/10.2174/1568026619666181130121558