A Deep Learning Framework for Robust and Accurate Prediction of ncRNA-Protein Interactions Using Evolutionary Information

Abstract

The interactions between non-coding RNAs (ncRNAs) and proteins play an important role in many biological processes, and their biological functions are primarily achieved by binding with a variety of proteins. High-throughput biological techniques are used to identify protein molecules bound with specific ncRNA, but they are usually expensive and time consuming. Deep learning provides a powerful solution to computationally predict RNA-protein interactions. In this work, we propose the RPI-SAN model by using the deep-learning stacked auto-encoder network to mine the hidden high-level features from RNA and protein sequences and feed them into a random forest (RF) model to predict ncRNA binding proteins. Stacked assembling is further used to improve the accuracy of the proposed method. Four benchmark datasets, including RPI2241, RPI488, RPI1807, and NPInter v2.0, were employed for the unbiased evaluation of five established prediction tools: RPI-Pred, IPMiner, RPISeq-RF, lncPro, and RPI-SAN. The experimental results show that our RPI-SAN model achieves much better performance than other methods, with accuracies of 90.77%, 89.7%, 96.1%, and 99.33%, respectively. It is anticipated that RPI-SAN can be used as an effective computational tool for future biomedical researches and can accurately predict the potential ncRNA-protein interacted pairs, which provides reliable guidance for biological research.

RPI-SE: a stacking ensemble learning framework for ncRNA-protein interactions prediction using sequence information

“…Finally, stacking ensemble is adopted to integrate these base predictors. To thoroughly verify the performance, the RPI-SE is evaluated on three benchmark data sets under fivefold cross-validation, including RPI369 [20], RPI488 [23] and RPI1807 [21], and compared with other methods, including RPISeq-RF [20], RPI-Pred [21], lncPro [14], IPMiner [23] and RPI-SAN [18]. The experimental results demonstrate that RPI-SE is competent for ncRPIs prediction task, obtained predictive performance with high accuracy and robustness.…”

“…High-throughput methods are valuable but timeconsuming and expensive. In recent years, there have been extensive research on computational prediction of proteins-RNAs interactions (RPIs) [14][15][16][17][18]. Pancaldi et al applied both Random Forest (RF) and Support Vector Machine (SVM) model for RPIs prediction, using more than 100 different functional and physical features, such as genomic context, structure or localization, experimental translation and so on [19].…”

“…In our previous work, we presented a deep learning stacked autoencoder network based framework to predict ncRNA-protein interactions, named RPI-SAN. The main contribution of RPI-SAN is the application of deep stacked autoencoder to obtain efficient hidden representation of RNA and protein sequence information [18,24]. Deep learning shows excellent ability with large-scale data support in many fields, however, ncRPIs data sets generally don't have large scales, thus it's not very suitable or urgent need for deep learning methods.…”

“…To this end, we propose a stacking ensemble based computational model, RPI-SE, by integrating Gradient Boosting Decision Tree (GBDT, implemented by XGBoost) [27], SVM [28,29] and Extremely randomized Trees [30] (ExtraTree) algorithms to predict ncRNA-protein interactions. Specifically, k-mer sparse matrix is used to exploit the sequence information of RNA, which retains not only the nucleic acid components, but also the sequence order information [18,31,32]. Meanwhile, the Legendre Moments (LMs) descriptor is applied to convert the information contained in a the Position Weight Matrix (PWM) [33,34] in a feature vector, which can retain the evolutionary information contained in amino acid sequences corresponding to physicochemical properties.…”

Construct a molecular associations network to systematically understand intermolecular associations inHumancells

“…Existing biomolecular interaction or association studies often focusing only on the links between individual molecules, including mRNA-protein interactions [1,2], lncRNA-protein interactions [3], protein-protein interactions [4], miRNA-protein interactions [5], miRNA-lncRNA interactions [6,7], considering exogenous chemical compound or complex disease, there is drug-protein interactions [8,9], drug-disease interactions [10][11][12], miRNA-disease associations [13,14], lncRNA-disease associations [15], protein-disease associations [10,16]. Emerging research on circRNA shows there are also circRNA-miRNA associations [17], circRNA-protein interactions [18] and circRNA-disease associations [19].…”