An improved interolog mapping-based computational prediction of protein-protein interactions with increased network...

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Identifying protein-protein interactions (PPIs) is crucial for understanding biological processes.
• Existing methods for interolog mapping (predicting interactions across organisms) require efficient and validated computational approaches.
• Public databases (pDBs) offer vast resources for PPI prediction but require robust analysis metrics.

Purpose of the Study:

• To compare six BLAST+ metrics for predicting PPIs using interolog mapping.
• To identify the most effective metric and assess the contribution of different public databases (String, IntAct, Psibase) in PPI prediction.
• To validate computational predictions against experimentally curated interaction data.

Main Methods:

• Interolog mapping of ortholog interactions from the Database of Interacting Proteins (DIP) to String, IntAct, and Psibase using BLAST+.
• Evaluation of six BLAST+ metrics (alignment score, e-value, bitscore, similarity, identity, coverage) and their combinations using Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC).
• Consideration of the top 20 BLAST+ hits and validation using a subset of DIP experimental interactions curated by IMEx.

Main Results:

• The best performing BLAST+ metric achieved an AUC of 0.96 for a single public database and 0.93 for combined databases.
• A cut-off point of 0.70 yielded high specificity (0.95) and sensitivity (0.90) for individual databases.
• The study identified the optimal metric and assessed the predictive power of individual and combined public databases.

Conclusions:

• The developed interolog mapping method, utilizing specific BLAST+ metrics, efficiently predicts protein-protein interactions.
• This approach offers a validated and efficient strategy for discovering novel PPIs in organisms of interest.
• The findings provide a robust computational tool for enhancing biological pathway and network analysis.