BgN-Score and BsN-Score: Bagging and boosting based ensemble neural networks scoring functions for accurate binding affinity prediction of protein-ligand complexes

Ashtawy, Hossam M.; Mahapatra, Nihar R.

doi:10.1186/1471-2105-16-s4-s8

Cited by 54 publications

(43 citation statements)

References 45 publications

(59 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Inspired by the outstanding performance of ensemble learning methods such as RF and boosted regression trees in a previous assessment study, Ashtawy and Mahapatra also proposed two ensemble NN‐based SFs based on bagging (BgN‐Score) and boosting (BsN‐Score). To make a better comparison, the same combinations of the features used in the previous assessment were used in this study.…”

Section: Traditional Machine Learning Methods In Scoring Functionsmentioning

confidence: 99%

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

Shen

Ding

Wang

et al. 2019

WIREs Comput Mol Sci

186

180

View full text Add to dashboard Cite

Molecule docking has been regarded as a routine tool for drug discovery, but its accuracy highly depends on the reliability of scoring functions (SFs). With the rapid development of machine learning (ML) techniques, ML-based SFs have gradually emerged as a promising alternative for protein-ligand binding affinity prediction and virtual screening, and most of them have shown significantly better performance than a wide range of classical SFs. Emergence of more data-hungry deep learning (DL) approaches in recent years further fascinates the exploitation of more accurate SFs. Here, we summarize the progress of traditional ML-based SFs in the last few years and provide insights into recently developed DL-based SFs. We believe that the continuous improvement in ML-based SFs can surely guide the early-stage drug design and accelerate the discovery of new drugs. This article is categorized under:Computer and Information Science > Chemoinformaticsdeep learning, machine learning, molecular docking, scoring function, structure-based drug design | INTRODUCTIONTraditional drug discovery largely relies on the application of high-throughput screening, an experimental technique with acceptable performance but high cost and low efficiency. 1 With the rapid development of computational chemistry and computer technology, computer-aided drug design (CADD) has gradually emerged as a powerful technique in the design and development of new drug candidates in the past three decades. 2 Virtual screening (VS), an important branch of CADD, can enrich potential actives from large virtual compound libraries through in silico methods rather than real experiments, which can not only accelerate the process of drug discovery but also greatly reduce the time and resource cost. 3-5 Depending on whether the three-dimensional (3D) structure of a target is used or not, VS approaches can be classified into two major categories: ligand-based virtual screening (LBVS) and structure-based virtual screening (SBVS). 6 LBVS aims to discover active molecules through the models developed based on a set of known ligands of a target of interest, which may limit its capability to find novel chemotypes. Compared with LBVS, SBVS is considered to be a better choice to discover novel active compounds if the 3D structure of a given target is available. 7 Chao Shen and Junjie Ding are equivalent first authors.

show abstract

Section: Traditional Machine Learning Methods In Scoring Functionsmentioning

confidence: 99%

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

Shen

Ding

Wang

et al. 2019

WIREs Comput Mol Sci

186

180

View full text Add to dashboard Cite

show abstract

“…The performance of the scoring functions was evaluated with RMSE, Pearson’s correlation coefficient (Rp), and Spearman’s rank correlation coefficient (Rs) between the predicted and measured binding affinities, because these are widely used to evaluate scoring functions [ 35 , 51 , 52 ]. RMSE represents the differences between the predicted and measured binding affinity values.…”

Section: Methodsmentioning

confidence: 99%

Virtual screening approach to identifying influenza virus neuraminidase inhibitors using molecular docking combined with machine-learning-based scoring function

Zhang

et al. 2017

Oncotarget

View full text Add to dashboard Cite

In recent years, an epidemic of the highly pathogenic avian influenza H7N9 virus has persisted in China, with a high mortality rate. To develop novel anti-influenza therapies, we have constructed a machine-learning-based scoring function (RF-NA-Score) for the effective virtual screening of lead compounds targeting the viral neuraminidase (NA) protein. RF-NA-Score is more accurate than RF-Score, with a root-mean-square error of 1.46, Pearson’s correlation coefficient of 0.707, and Spearman’s rank correlation coefficient of 0.707 in a 5-fold cross-validation study. The performance of RF-NA-Score in a docking-based virtual screening of NA inhibitors was evaluated with a dataset containing 281 NA inhibitors and 322 noninhibitors. Compared with other docking–rescoring virtual screening strategies, rescoring with RF-NA-Score significantly improved the efficiency of virtual screening, and a strategy that averaged the scores given by RF-NA-Score, based on the binding conformations predicted with AutoDock, AutoDock Vina, and LeDock, was shown to be the best strategy. This strategy was then applied to the virtual screening of NA inhibitors in the SPECS database. The 100 selected compounds were tested in an in vitro H7N9 NA inhibition assay, and two compounds with novel scaffolds showed moderate inhibitory activities. These results indicate that RF-NA-Score improves the efficiency of virtual screening for NA inhibitors, and can be used successfully to identify new NA inhibitor scaffolds. Scoring functions specific for other drug targets could also be established with the same method.

show abstract

“…Machine learning methods at varying levels of sophistication have long been considered in the context of structure-based virtual screening [39,31,32,[40][41][42][43][44][45][46]29,[47][48][49][50][51][52][53][54]. The vast majority of such studies sought to train a regression model that would recapitulate the binding affinities of known complexes, and thus provide a natural and intuitive replacement for traditional scoring functions [31,32,[41][42][43][44][45][46]29,47,49,[51][52][53][54]. The downside of such a strategy, however, is that the resulting models are not ever exposed to any inactive complexes in the course of training: this is especially important in the context of docked complexes arising from virtual screening, where most compounds in the library are presumably inactive.…”

Section: Developing a Challenging Training Setmentioning

confidence: 99%

Machine learning classification can reduce false positives in structure-based virtual screening

Adeshina

Deeds

Karanicolas

2020

Preprint

View full text Add to dashboard Cite

With the recent explosion in the size of libraries available for screening, virtual screening is positioned to assume a more prominent role in early drug discovery's search for active chemical matter.Modern virtual screening methods are still, however, plagued with high false positive rates: typically, only about 12% of the top-scoring compounds actually show activity when tested in biochemical assays.We argue that most scoring functions used for this task have been developed with insufficient thoughtfulness into the datasets on which they are trained and tested, leading to overly simplistic models and/or overtraining. These problems are compounded in the literature because none of the studies reporting new scoring methods have validated their model prospectively within the same study. Here, we report a new strategy for building a training dataset (D-COID) that aims to generate highly-compelling decoy complexes that are individually matched to available active complexes. Using this dataset, we train a general-purpose classifier for virtual screening (vScreenML) that is built on the XGBoost framework of gradient-boosted decision trees. In retrospective benchmarks, our new classifier shows outstanding performance relative to other scoring functions. We additionally evaluate the classifier in a prospective context, by screening for new acetylcholinesterase inhibitors. Remarkably, we find that nearly all compounds selected by vScreenML show detectable activity at 50 µM, with 10 of 23 providing greater than 50% inhibition at this concentration. Without any medicinal chemistry optimization, the most potent hit from this initial screen has an IC50 of 280 nM, corresponding to a Ki value of 173 nM. These results support using the D-COID strategy for training classifiers in other computational biology tasks, and for vScreenML in virtual screening campaigns against other protein targets. Both D-COID and vScreenML are freely distributed to facilitate such efforts.

show abstract

BgN-Score and BsN-Score: Bagging and boosting based ensemble neural networks scoring functions for accurate binding affinity prediction of protein-ligand complexes

Cited by 54 publications

References 45 publications

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

From machine learning to deep learning: Advances in scoring functions for protein–ligand docking

Virtual screening approach to identifying influenza virus neuraminidase inhibitors using molecular docking combined with machine-learning-based scoring function

Machine learning classification can reduce false positives in structure-based virtual screening

Contact Info

Product

Resources

About