A deep learning framework for identifying essential proteins based on multiple biological information

Abstract

Background: Essential Proteins are demonstrated to exert vital functions on cellular processes and are indispensable for the survival and reproduction of the organism. Traditional centrality methods perform poorly on complex protein-protein interaction (PPI) networks. Machine learning approaches based on high-throughput data lack the exploitation of the temporal and spatial dimensions of biological information.

Results: We put forward a deep learning framework to predict essential proteins by integrating features obtained from the PPI network, subcellular localization, and gene expression profiles. In our model, the node2vec method is applied to learn continuous feature representations for proteins in the PPI network, which capture the diversity of connectivity patterns in the network. The concept of depthwise separable convolution is employed on gene expression profiles to extract properties and observe the trends of gene expression over time under different experimental conditions. Subcellular localization information is mapped into a long one-dimensional vector to capture its characteristics. Additionally, we use a sampling method to mitigate the impact of imbalanced learning when training the model. With experiments carried out on the data of Saccharomyces cerevisiae, results show that our model outperforms traditional centrality methods and machine learning methods. Likewise, the comparative experiments have manifested that our process of various biological information is preferable.

Conclusions: Our proposed deep learning framework effectively identifies essential proteins by integrating multiple biological data, proving a broader selection of subcellular localization information significantly improves the results of prediction and depthwise separable convolution implemented on gene expression profiles enhances the performance.

Keywords: Deep learning; Essential protein; Gene expression; Protein–protein interaction network; Subcellular localization.

Fig. 1

General structure of our framework. Conv1D 1-dimensional convolution, BN batch normalization, MP max pooling, PW pointwise convolution, GMP global max pooling

Fig. 2

The PCC between replicate samples and different conditions in the gene expression profiles. C control group, O observation group, R1 replicate sample 1, R2 replicate sample 2, R3 replicate sample 3

Fig. 3

Visualization of gene expression profile processing. Conv1D 1-dimensional convolution, BN batch normalization, MP max pooling, PW pointwise convolution, GMP global max pooling

Fig. 4

The process of subcellular localization data for protein YLR308W. ABmEE complex Apolipoprotein B mRNA editing enzyme complex, CPKAK complex cyclin-dependent protein kinase activating kinase holoenzyme complex, 6PFK complex 6-phospho fructose kinase complex

Fig. 5

ROC and PR curve of different processes on gene expression profiles

Fig. 6

ROC and PR curve of different selections of subcellular localization

Fig. 7

ROC and PR curve of node2vec technique and centrality methods

Fig. 8

ROC and PR curve of different feature combinations. S subcellular localization features, N network embedding features, G gene expression profile features, N + G network embedding features plus gene expression profile features, N + S network embedding features plus subcellular localization features, S + G subcellular localization features plus gene expression profile features, S + N + G subcellular localization features plus network embedding features and gene expression profile features

Fig. 9

ROC and PR curve of different approaches to the unbalanced dataset. CW class weight

Fig. 10

Performance of our method and centrality methods

Fig. 11

ROC and PR curve of our method and other machine learning methods. SVM support vector machine, LR logistic regression, NB naive bayes, RF random forest, DT decision tree

Fig. 12

The process of GSE41828

Fig. 13

Compare with centrality methods on Homo sapiens data

Fig. 14

ROC and PR curve of our method and other machine learning methods. SVM Support Vector Machine, AB AdaBoost, LR Logistic Regression, NB Naive Bayes, RF Random Forest, DT Decision Tree