ET‐score: Improving Protein‐ligand Binding Affinity Prediction Based on Distance‐weighted Interatomic Contact Features Using Extremely Randomized Trees Algorithm

Abstract

The molecular docking simulation is a key computational tool in modern drug discovery research that its predictive performance strongly depends on the employed scoring functions. Many recent studies have shown that the application of machine learning algorithms in the development of scoring functions has led to a significant improvement in docking performance. In this work, we introduce a new machine learning (ML) based scoring function called ET-Score, which employs the distanceweighted interatomic contacts between atom type pairs of the ligand and the protein for featurizing protein À ligand complexes and Extremely Randomized Trees algorithm for the training process. The performance of ET-Score is compared with some successful ML-based scoring functions and several popular classical scoring functions on the PDBbind 2016v core set. It is shown that our ET-Score model (with Pearson's correlation of 0.827 and RMSE of 1.332) achieves very good performance in comparison with most of the ML-based scoring functions and all classical scoring functions despite its extremely low computational cost. ET-Score's codes are freely available on the web at https://github.com/miladrayka/ET_Score.

Protein–Ligand Docking in the Machine-Learning Era

“…Recently, customized protein–ligand interaction features became popular in scoring function development, such as ET-score (2021) and ECIF-GBT (2021) [ 104 , 106 ]. ET-score employed protein–ligand interaction features defined by distance-weighted interatomic contacts between atom type pairs of the protein and ligand.…”

GB‐score: Minimally designed machine learning scoring function based on distance‐weighted interatomic contact features

“…As mentioned before, after excluding the core set, the refined set 2016, the composition of the refined and the general sets 2016, and the same composition for 2019 are used as the training sets. The distance‐weighted interatomic contact featurization method was applied to protein‐ligand complexes to generate a numerical representation for them [24]. RF, ET, and GBT were adopted as fast and standard learning algorithms to discern hidden patterns in the training data.…”

“…Distances with magnitude below the predefined cutoff ( d cutoff ) are weighted by an inverse power of a natural number ( n ) and sum together. In our previous work, we demonstrated that 12 $${\dot{A}}$$ and 2 are appropriate choices for d cutoff and n , respectively [24]. The mentioned algorithm is repeated iteratively for all possible atom types pairs, and a feature vector with 400 dimensions as a representation of a protein‐ligand complex is produced [24]: $$$\vcenter{\openup.5em\halign{$\displaystyle{#}$\cr \vec{X}=\left\{{X}_{H,{H}_{p}},{X}_{H,{C}_{p}},\dots, {X}_{I,{I}_{h}}\right\}\hfill\cr}}$$$ $$$\vcenter{\openup.5em\halign{$\displaystyle{#}$\cr {X}_{i,j}=\sum _{k=1}^{{K}_{j}}\sum _{l=1}^{{L}_{i}}{{1}\over{{d}_{lk}^{2}}}\hfill\cr}}$$$ …”

LCP: Simple Representation of Docking Poses for Machine Learning: A Case Study on Xanthine Oxidase Inhibitors

“…Additionally, the establishment of an in‐silico molecular design approach is still required [2–5] . Compared to ligand‐based drug design (LBDD), structure‐based drug design (SBDD), which is closely related to this study, is significantly advantageous in terms of identifying active compounds with novel molecular frameworks and different chemical classes [6,7] . Currently, many clinical drugs, including kinase inhibitors, such as erdafitinib, pexidartinib, and vemurafenib, have been developed through SBDD, [8] which has emerged as an effective approach for drug discovery.…”

“…[2][3][4][5] Compared to ligand-based drug design (LBDD), structure-based drug design (SBDD), which is closely related to this study, is significantly advantageous in terms of identifying active compounds with novel molecular frameworks and different chemical classes. [6,7] Currently, many clinical drugs, including kinase inhibitors, such as erdafitinib, pexidartinib, and vemurafenib, have been developed through SBDD, [8] which has emerged as an effective approach for drug discovery. However, one of the challenges associated with this approach is the difficulty in obtaining accurate ligandbinding poses.…”