Machine‐learning scoring functions to improve structure‐based binding affinity prediction and virtual screening

Abstract

Docking tools to predict whether and how a small molecule binds to a target can be applied if a structural model of such target is available. The reliability of docking depends, however, on the accuracy of the adopted scoring function (SF). Despite intense research over the years, improving the accuracy of SFs for structure‐based binding affinity prediction or virtual screening has proven to be a challenging task for any class of method. New SFs based on modern machine‐learning regression models, which do not impose a predetermined functional form and thus are able to exploit effectively much larger amounts of experimental data, have recently been introduced. These machine‐learning SFs have been shown to outperform a wide range of classical SFs at both binding affinity prediction and virtual screening. The emerging picture from these studies is that the classical approach of using linear regression with a small number of expert‐selected structural features can be strongly improved by a machine‐learning approach based on nonlinear regression allied with comprehensive data‐driven feature selection. Furthermore, the performance of classical SFs does not grow with larger training datasets and hence this performance gap is expected to widen as more training data becomes available in the future. Other topics covered in this review include predicting the reliability of a SF on a particular target class, generating synthetic data to improve predictive performance and modeling guidelines for SF development. WIREs Comput Mol Sci 2015, 5:405–424. doi: 10.1002/wcms.1225For further resources related to this article, please visit the WIREs website.

Automatic selection of molecular descriptors using random forest: Application to drug discovery

“…Both SBVS and LBVS methods use different forms of scoring functions for affinity prediction and can complement high-throughput screening techniques; however, accurate prediction of binding affinity by any vir-15 tual screening method is a very challenging task. Use of modern computational intelligence techniques that do not impose a pre-determined scoring function has generated interest as a mean to improve prediction accuracy (Ain et al, 2015;Ballester & Mitchell, 2010). Selection of chemical characteristics (molecular descriptors) with greater discriminatory power has the potential to improve 20 scoring predictions of which compounds will be good candidates, i.e., bioactive.…”

Recent advances in automated protein design and its future challenges

“…Machine-learning-based scoring functions uses a variety of mostly supervised machine-learning algorithms [158,159], such as artificial neural networks [160], random forest [161–163], and support vector machine [164], to learn about a specific energetic or other structural or biological properties using a training set of protein structures. The resulting, trained, machine-learning-based function, can then be used to produce a scoring value associated with a predicted property: $$\begin{gathered} \mathbf{Input}:\text{Descriptors}\to \text{Trained}\text{Scoring}\text{Function}\to \mathbf{Output} \\ :\text{Scoring}\text{Value} \end{gathered}$$…”

An Efficient ABC_DE_Based Hybrid Algorithm for Protein–Ligand Docking

“…A scoring function [11,12,13,14,15,16] and a search algorithm [17,18,19] are the necessary tools of a docking method for solving the two goals above. The scoring function is used to evaluate the affinity between the receptor and the ligand for each conformation [20].…”