Dataset Information

Explainable Deep Relational Networks for Predicting Compound-Protein Affinities and Contacts.

ABSTRACT: Predicting compound-protein affinity is beneficial for accelerating drug discovery. Doing so without the often-unavailable structure data is gaining interest. However, recent progress in structure-free affinity prediction, made by machine learning, focuses on accuracy but leaves much to be desired for interpretability. Defining intermolecular contacts underlying affinities as a vehicle for interpretability; our large-scale interpretability assessment finds previously used attention mechanisms inadequate. We thus formulate a hierarchical multiobjective learning problem, where predicted contacts form the basis for predicted affinities. We solve the problem by embedding protein sequences (by hierarchical recurrent neural networks) and compound graphs (by graph neural networks) with joint attentions between protein residues and compound atoms. We further introduce three methodological advances to enhance interpretability: (1) structure-aware regularization of attentions using protein sequence-predicted solvent exposure and residue-residue contact maps; (2) supervision of attentions using known intermolecular contacts in training data; and (3) an intrinsically explainable architecture where atomic-level contacts or "relations" lead to molecular-level affinity prediction. The first two and all three advances result in DeepAffinity+ and DeepRelations, respectively. Our methods show generalizability in affinity prediction for molecules that are new and dissimilar to training examples. Moreover, they show superior interpretability compared to state-of-the-art interpretable methods: with similar or better affinity prediction, they boost the AUPRC of contact prediction by around 33-, 35-, 10-, and 9-fold for the default test, new-compound, new-protein, and both-new sets, respectively. We further demonstrate their potential utilities in contact-assisted docking, structure-free binding site prediction, and structure-activity relationship studies without docking. Our study represents the first model development and systematic model assessment dedicated to interpretable machine learning for structure-free compound-protein affinity prediction.

SUBMITTER: Karimi M

PROVIDER: S-EPMC7987499 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundThe ability to predict which pairs of amino acid residues in a protein are in contact with each other offers many advantages for various areas of research that focus on proteins. For example, contact prediction can be used to reduce the computational complexity of predicting the structure of proteins and even to help identify functionally important regions of proteins. These predictions are becoming especially important given the relatively low number of experimentally determined protein structures compared to the amount of available protein sequence data.ResultsHere we have developed and benchmarked a set of machine learning methods for performing residue-residue contact prediction, including random forests, direct-coupling analysis, support vector machines, and deep networks (stacked denoising autoencoders). These methods are able to predict contacting residue pairs given only the amino acid sequence of a protein. According to our own evaluations performed at a resolution of +/- two residues, the predictors we trained with the random forest algorithm were our top performing methods with average top 10 prediction accuracy scores of 85.13% (short range), 74.49% (medium range), and 54.49% (long range). Our ensemble models (stacked denoising autoencoders combined with support vector machines) were our best performing deep network predictors and achieved top 10 prediction accuracy scores of 75.51% (short range), 60.26% (medium range), and 43.85% (long range) using the same evaluation. These tests were blindly performed on targets from the CASP11 dataset; and the results suggested that our models achieved comparable performance to contact predictors developed by groups that participated in CASP11.ConclusionsDue to the challenging nature of contact prediction, it is beneficial to develop and benchmark a variety of different prediction methods. Our work has produced useful tools with a simple interface that can provide contact predictions to users without requiring a lengthy installation process. In addition to this, we have released our C++ implementation of the direct-coupling analysis method as a standalone software package. Both this tool and our RFcon web server are freely available to the public at http://dna.cs.miami.edu/RFcon /.

Project description:Machine learning methods for protein function prediction are urgently needed, especially now that a substantial fraction of known sequences remains unannotated despite the extensive use of functional assignments based on sequence similarity. One major bottleneck supervised learning faces in protein function prediction is the structured, multi-label nature of the problem, because biological roles are represented by lists of terms from hierarchically organised controlled vocabularies such as the Gene Ontology. In this work, we build on recent developments in the area of deep learning and investigate the usefulness of multi-task deep neural networks (MTDNN), which consist of upstream shared layers upon which are stacked in parallel as many independent modules (additional hidden layers with their own output units) as the number of output GO terms (the tasks). MTDNN learns individual tasks partially using shared representations and partially from task-specific characteristics. When no close homologues with experimentally validated functions can be identified, MTDNN gives more accurate predictions than baseline methods based on annotation frequencies in public databases or homology transfers. More importantly, the results show that MTDNN binary classification accuracy is higher than alternative machine learning-based methods that do not exploit commonalities and differences among prediction tasks. Interestingly, compared with a single-task predictor, the performance improvement is not linearly correlated with the number of tasks in MTDNN, but medium size models provide more improvement in our case. One of advantages of MTDNN is that given a set of features, there is no requirement for MTDNN to have a bootstrap feature selection procedure as what traditional machine learning algorithms do. Overall, the results indicate that the proposed MTDNN algorithm improves the performance of protein function prediction. On the other hand, there is still large room for deep learning techniques to further enhance prediction ability.

Dataset Information

Explainable Deep Relational Networks for Predicting Compound-Protein Affinities and Contacts.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets