Development and evaluation of a deep learning model for protein–ligand binding affinity prediction

Stepniewska-Dziubinska, Marta M.; Zielenkiewicz, Piotr; Siedlecki, Paweł

doi:10.1093/bioinformatics/bty374

Cited by 479 publications

(604 citation statements)

References 50 publications

Supporting

Mentioning

553

Contrasting

Unclassified

Order By: Relevance

“…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

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

show abstract

“…Similarly to recent works addressing pocket detection (Jiménez et al, 2017) and protein-ligand affinity prediction (Jiménez et al, 2018;Stepniewska-Dziubinska et al, 2018), we regard protein structures as 3D images with c channels f : R 3 → R c (4D tensors). This is analogous to the treatment of color images in computer vision as functions assigning a vector of intensities of three primary colors to each pixel, R 2 → R 3 .…”

Section: Volumetric Input Representationmentioning

confidence: 99%

“…This trend has also reached the community of structural biology and computational chemistry, showing utility in a range of scenarios relevant to drug discovery (Rifaioglu et al, 2018). In particular, trainable methods have been applied to protein structure data for a number of applications including protein-ligand affinity prediction (Wallach et al, 2015;Gomes et al, 2017;Ragoza et al, 2017;Stepniewska-Dziubinska et al, 2018;Jiménez et al, 2018;Imrie et al, 2018), protein structure prediction (AlQuraishi, 2018;Evans et al, 2018), binding pocket inpainting (Škalič et al, 2019), binding site detection (Jiménez et al, 2017) and prediction of protein-protein interaction sites (Fout et al, 2017;Townshend et al, 2018). However, to our knowledge a deep learning approach to pocket matching has not been previously described.…”

Section: Introductionmentioning

confidence: 99%

DeeplyTough: Learning Structural Comparison of Protein Binding Sites

Simonovsky

Meyers

2019

Preprint

View full text Add to dashboard Cite

Motivation: Protein binding site comparison (pocket matching) is of importance in drug discovery. Identification of similar binding sites can help guide efforts for hit finding, understanding polypharmacology and characterization of protein function. The design of pocket matching methods has traditionally involved much intuition, and has employed a broad variety of algorithms and representations of the input protein structures. We regard the high heterogeneity of past work and the recent availability of large-scale benchmarks as an indicator that a data-driven approach may provide a new perspective. Results: We propose DeeplyTough, a convolutional neural network that encodes a three-dimensional representation of protein binding sites into descriptor vectors that may be compared efficiently in an alignment-free manner by computing pairwise Euclidean distances. The network is trained with supervision: (i) to provide similar pockets with similar descriptors, (ii) to separate the descriptors of dissimilar pockets by a minimum margin, and (iii) to achieve robustness to nuisance variations. We evaluate our method using three large-scale benchmark datasets, on which it demonstrates excellent performance for held-out data coming from the training distribution and competitive performance when the trained network is required to generalize to datasets constructed independently.

show abstract

“…Deep learning has been relatively widely used by the bioinformatics [21] and computational biology. [20] Deep neural network models have been developed for prediction protein-binding sites, [23,24] predicting protein DNA and RNA binding sites, [25][26][27][28] protein secondary structure predictions, [29,30] protein fold recognition, [31] protein-ligand binding affinity predictions, [32] prediction of proteinÀ protein interactions, [33] predict protein folding, [34] protein contact map predictions, [35] predict splicing patterns, [36] peptide-MHC binding predictions, [37] liver injury predictions, [38] regulatory genomics, [39] learning the functional activities of DNA sequences from genomics data. [40] In this work, we propose three different deep learning approaches which could generate a highly accurate prediction of protein metal-binding sites and show that they are able to capture binding site characteristics and can outperform the benchmark algorithm.…”

Section: Introductionmentioning

confidence: 99%

Prediction of Protein Metal Binding Sites Using Deep Neural Networks

Haberal

Ogul

2019

Molecular Informatics

View full text Add to dashboard Cite

Metals have crucial roles for many physiological, pathological and diagnostic processes. Metal binding proteins or metalloproteins are important for metabolism functions. The proteins that reach the three-dimensional structure by folding show which vital function is fulfilled. The prediction of metal-binding in proteins will be considered as a step-in function assignment for new proteins, which helps to obtain functional proteins in genomic studies, is critical to protein function annotation and drug discovery. Computational predictions made by using machine learning methods from the data obtained from amino acid sequences are widely used in the protein metalbinding and various bioinformatics fields. In this work, we present three different deep learning architectures for prediction of metal-binding of Histidines (HIS) and Cysteines (CYS) amino acids. These architectures are as follows: 2D Convolutional Neural Network, Long-Short Term Memory and Recurrent Neural Network. Their comparison is carried out on the three different sets of attributes derived from a public dataset of protein sequences. These three sets of features extracted from the protein sequence were obtained using the PAM scoring matrix, protein composition server, and binary representation methods. The results show that a better performance for prediction of protein metal-binding sites is obtained through Convolutional Neural Network architecture.

show abstract

Development and evaluation of a deep learning model for protein–ligand binding affinity prediction

Cited by 479 publications

References 50 publications

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

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

DeeplyTough: Learning Structural Comparison of Protein Binding Sites

Prediction of Protein Metal Binding Sites Using Deep Neural Networks

Contact Info

Product

Resources

About