Decision Tree for Prediction of Binding Affinity

Recent advances in machine learning methods indicate the adequacy of these approaches to build scoring functions to predict binding affinity. Applying the scoring function to determine protein-ligand interaction is a pivotal step in the early stages of drug discovery projects, where we integrate docking results and machine learning techniques to create an adequate model for a protein system of interest. Here, we focus on regression models built using Decision Trees to address protein-ligand interactions. This powerful machine learning technique can build models to address complex systems for classification and regression problems. We show how to apply the Decision Tree method to explore the concept of scoring function space using the program SKReg4Model. This program builds regression models using features from Vina Force Field, and energy terms are determined using Molegro Virtual Docker or any docking program. We based its code on the program SAnDReS 2.0 and the Scikit-Learn library. All datasets and a Jupyter Notebook with SKReg4Model discussed in this work are available at GitHub: https://github.com/azevedolab/docking#readme . We made the program SAnDReS 2.0 available at https://github.com/azevedolab/sandres .