Essential Protein Prediction Based on node2vec and XGBoost.

Area of Science:

• Computational biology
• Systems biology
• Bioinformatics

Background:

• Essential proteins are critical for organism and cellular survival.
• Identifying essential proteins aids in understanding protein functions and discovering drug targets.
• Traditional experimental methods for identification are costly and time-consuming, necessitating computational approaches.

Purpose of the Study:

• To propose a novel computational model, Ess-NEXG, for accurate essential protein identification.
• To address the challenge of noise in protein-protein interaction (PPI) networks affecting prediction accuracy.
• To integrate multiple biological information sources for constructing a more credible PPI network.

Main Methods:

• Constructed a credible weighted PPI network by integrating orthologous, subcellular localization, and RNA-Seq information.
• Utilized the node2vec technique to extract topological features of proteins within the weighted PPI network.
• Employed the eXtreme Gradient Boosting (XGBoost) algorithm for essential protein prediction based on extracted features.

Main Results:

• The proposed Ess-NEXG model demonstrated superior performance compared to existing computational methods.
• Integration of diverse biological data effectively reduced noise in the PPI network.
• The node2vec and XGBoost combination proved effective in leveraging topological features for prediction.

Conclusions:

• Ess-NEXG offers a more accurate and efficient approach to essential protein identification.
• The method highlights the importance of integrating multi-source biological data for robust network construction.
• This study provides a valuable tool for biological research and drug discovery.