EPI-SF: essential protein identification in protein interaction networks using sequence features.

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning Applications in Proteomics

Background:

• Identifying essential proteins is crucial for understanding organism viability and function.
• Traditional methods for protein identification are time-consuming, labor-intensive, and costly.
• Computational approaches, particularly machine learning, offer a more efficient alternative.

Purpose of the Study:

• To propose and evaluate a novel machine learning-based technique, EPI-SF, for identifying essential proteins.
• To compare the performance of various machine learning models using protein sequence features.
• To apply the validated method to predict essential proteins in human protein-protein interaction networks (PPINs) and assess their potential involvement in diseases like COVID-19.

Main Methods:

• Extraction of protein sequence features from protein-protein interaction networks (PPINs).
• Application of conventional machine learning (ML) models including XGBoost, AdaBoost, Logistic Regression, SVM, Decision Tree, Random Forest, and Naïve Bayes.
• Evaluation of model performance using metrics such as precision, recall, F1-score, and AUC, with a focus on the Random Forest model.

Main Results:

• The Random Forest model achieved the highest performance in identifying essential proteins in yeast PPINs, with precision, recall, F1-score, and AUC values of 0.703, 0.720, 0.711, and 0.745, respectively.
• EPI-SF, utilizing the Random Forest model, outperformed traditional centrality measures (e.g., betweenness centrality, closeness centrality) and deep learning methods (e.g., DeepEP).
• The method was successfully applied to predict novel essential proteins in the human PPIN.

Conclusions:

• The proposed EPI-SF technique, powered by machine learning and protein sequence features, is an effective approach for identifying essential proteins.
• The Random Forest model demonstrates significant potential for essential protein prediction compared to existing methods.
• The findings suggest potential roles for identified essential proteins in disease transmission, including COVID-19, warranting further investigation.