DeepDISOBind: accurate prediction of RNA-, DNA- and protein-binding intrinsically disordered residues with deep multi-task learning

Abstract

Proteins with intrinsically disordered regions (IDRs) are common among eukaryotes. Many IDRs interact with nucleic acids and proteins. Annotation of these interactions is supported by computational predictors, but to date, only one tool that predicts interactions with nucleic acids was released, and recent assessments demonstrate that current predictors offer modest levels of accuracy. We have developed DeepDISOBind, an innovative deep multi-task architecture that accurately predicts deoxyribonucleic acid (DNA)-, ribonucleic acid (RNA)- and protein-binding IDRs from protein sequences. DeepDISOBind relies on an information-rich sequence profile that is processed by an innovative multi-task deep neural network, where subsequent layers are gradually specialized to predict interactions with specific partner types. The common input layer links to a layer that differentiates protein- and nucleic acid-binding, which further links to layers that discriminate between DNA and RNA interactions. Empirical tests show that this multi-task design provides statistically significant gains in predictive quality across the three partner types when compared to a single-task design and a representative selection of the existing methods that cover both disorder- and structure-trained tools. Analysis of the predictions on the human proteome reveals that DeepDISOBind predictions can be encoded into protein-level propensities that accurately predict DNA- and RNA-binding proteins and protein hubs. DeepDISOBind is available at https://www.csuligroup.com/DeepDISOBind/

Complementarity of the residue-level protein function and structure predictions in human proteins

“…The results that we report in Section 3.1 reveal that DisoRDPbind secures AUC = 0.65 for the DNA binding and AUC = 0.62 for the RNA binding. This is consistent with a recent assessment where DisoRDPbind’s AUC are 0.67 and 0.60, respectively [92] . Similarly, we report AUC = 0.68 for DNA binding and AUC = 0.60 for RNA binding for DRNApred, while the previously published results are 0.68 and 0.65, respectively [89] .…”

Computational tools to study RNA-protein complexes

“…These methods extract various features from amino acid residues and use them as the input to train the machine-learning models for classification. Several algorithms like SVM (Cai et al, 2003;Kumar et al, 2008;Murakami et al, 2010;Wang et al, 2010;Walia et al, 2014;Yang et al, 2015a;Bressin et al, 2019;Su et al, 2019;Qiu et al, 2020), neural networks (Alipanahi et al, 2015;Peng et al, 2017;Yan and Kurgan, 2017;Deng et al, 2018;Zhao and Du, 2020;Zhang et al, 2021b;Sun et al, 2021;Zhang et al, 2022), naive bayes classifier (Sharan et al, 2017;Deng et al, 2021; etc, as listed in Supplementary Table S1, have been successfully implemented. A common limitation faced by such ML-based methods is that the extracted features may be poorly representative of the physicochemical and environmental properties of amino acid residues, or their simplistic combination may introduce redundancy and affect overall prediction power of the approaches.…”

Prediction of protein-protein interaction sites in intrinsically disordered proteins

“…Neural network is widely used in the study of protein-protein interaction. In this paper, we introduce five types of neural network models: Feedforward neural network (Zell, 1994), Bidirectional recurrent neural network (BRNN) (Mooney et al, 2012), Two-hidden layer neural network (Faraggi et al, 2009), Long Short-Term Memory (LSTM) (Hochreiter and Schmidhuber, 1997), and multi-task deep neural network (Zhang et al, 2021).…”

“…Based on the information derived from protein sequences, DeepDISOBind (Zhang et al, 2021) applied multi-task deep neural network to accurately predict the binding regions of disordered protein with DNA, RNA and protein. DeepDISOBind included shared layer, nucleic acid binding layer, protein binding layer, DNA binding layer and RNA binding layer.…”