ReSAPP: predicting overlapping protein complexes by merging multiple-sampled partitions of proteins

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Proteins function in complexes, necessitating accurate identification from protein-protein interaction (PPI) data.
• Predicting protein complexes from large-scale PPI networks is a significant challenge in systems biology.

Purpose of the Study:

• To introduce a novel computational method, Repeated Simulated Annealing of Partitions of Proteins (ReSAPP), for predicting protein complexes from weighted PPIs.
• To evaluate the performance of ReSAPP against existing state-of-the-art methods.

Main Methods:

• ReSAPP employs a two-stage approach: repeated simulated annealing to generate multiple protein partitions, followed by merging clusters of size two or more.
• The method utilizes weighted PPI data as input.

Main Results:

• ReSAPP demonstrated superior performance compared to established tools like MCL, MCODE, DPClus, CMC, COACH, RRW, NWE, and PPSampler.
• The F-measure of ReSAPP was significantly higher than a variant lacking the partition merging step, highlighting the effectiveness of the combined strategy.

Conclusions:

• The combination of sampling multiple partitions and merging them is an effective strategy for predicting protein complexes.
• ReSAPP offers an improved computational approach for identifying functional protein groups from PPI networks.