Comprehensive strategies of machine-learning-based quantitative structure-activity relationship models

Abstract

Summary Early quantitative structure-activity relationship (QSAR) technologies have unsatisfactory versatility and accuracy in fields such as drug discovery because they are based on traditional machine learning and interpretive expert features. The development of Big Data and deep learning technologies significantly improve the processing of unstructured data and unleash the great potential of QSAR. Here we discuss the integration of wet experiments (which provide experimental data and reliable verification), molecular dynamics simulation (which provides mechanistic interpretation at the atomic/molecular levels), and machine learning (including deep learning) techniques to improve QSAR models. We first review the history of traditional QSAR and point out its problems. We then propose a better QSAR model characterized by a new iterative framework to integrate machine learning with disparate data input. Finally, we discuss the application of QSAR and machine learning to many practical research fields, including drug development and clinical trials.

A Deep Learning-Based Quantitative Structure–Activity Relationship System Construct Prediction Model of Agonist and Antagonist with High Performance

“…Quantitative structure-activity relationship (QSAR) models can reduce the time and cost of molecular screening through mathematical prediction models of regression or classification of properties and activities of a chemical compound based on their chemical structure and statistically significant corresponding physicochemical/toxicological properties with other methods such as homology modeling, molecular docking, and molecular dynamics (MD) simulation . The structure-based molecular design mainly includes a receptor-based method through a three-dimensional (3D) chemical structure to obtain ligand interaction [1,35,36]. However, traditional QSAR models may frequently miss suitable candidate molecules, because of the poor predictive accuracy and versatility caused by poor feature selection that requires skill and knowledge and conformational limitations for coincidence effect [1,[37][38][39].…”

“…The structure-based molecular design mainly includes a receptor-based method through a three-dimensional (3D) chemical structure to obtain ligand interaction [1,35,36]. However, traditional QSAR models may frequently miss suitable candidate molecules, because of the poor predictive accuracy and versatility caused by poor feature selection that requires skill and knowledge and conformational limitations for coincidence effect [1,[37][38][39]. Therefore, a QSAR system with high-throughput and performance is desired because of the development of novel medicines, chemicals, and nanomaterials on human health.…”

“…The important factor for solving the QSAR issue is the extraction of information-rich numerical molecular descriptors associated with physicochemical/toxicological properties. However, 3D-QSAR has a high computational cost, and its performance is sensitive to changes in the ligand geometry such as conformation and orientation [1,40]. To resolve these drawbacks, 4D-QSAR, called MD-QSAR, applied the ligand geometry problem for the effective ligand constrains using a 4D-chemical descriptor with multiple structural conformation, orientation, and protonation state calculated through short run MD stimulation to approximate the Boltzman sampling [1,41,42].…”

ER/AR Multi-Conformational Docking Server: A Tool for Discovering and Studying Estrogen and Androgen Receptor Modulators

“…The in silico methods, especially the ligand-based QSAR (quantitative structure–activity relationships) approaches and structural base docking methods, are widely used in the computer-aided drug design field ( Mao et al, 2021 ; Sabe et al, 2021 ). These methods are applied to predict the compound’s activities against endocrine-related proteins, such as the estrogen receptor (ER) and androgen receptor (AR) ( Schneider et al, 2019 ).…”

Machine Learning Analysis of Essential Oils from Cuban Plants: Potential Activity against Protozoa Parasites

“…A decision tree establishes rules for decision making, that is, the algorithm will generate a structure like a flowchart with "nodes" where a condition will be checked and if met, the flow follows one branch, otherwise, it follows another, always leading to the nearest "node" where further decisionmaking will take place, until the end of the tree. Thus, given a training set, the algorithm will analyze the data and look for the best conditions and where to insert each data into the flow [25][26][27][28].…”