A comparative analysis of computational approaches and algorithms for protein subcomplex identification

Nazar Zaki, Antonio Mora

    Research output: Contribution to journalArticlepeer-review

    14 Citations (Scopus)

    Abstract

    High-throughput AP-MS methods have allowed the identification of many protein complexes. However, most post-processing methods of this type of data have been focused on detection of protein complexes and not its subcomplexes. Here, we review the results of some existing methods that may allow subcomplex detection and propose alternative methods in order to detect subcomplexes from AP-MS data. We assessed and drew comparisons between the use of overlapping clustering methods, methods based in the core-attachment model and our own prediction strategy (TRIBAL). The hypothesis behind TRIBAL is that subcomplex-building information may be concealed in the multiple edges generated by an interaction repeated in different contexts in raw data. The CACHET method offered the best results when the evaluation of the predicted subcomplexes was carried out using both the hypergeometric and geometric scores. TRIBAL offered the best performance when using a strict meet-min score.

    Original languageEnglish
    Article number4262
    JournalScientific reports
    Volume4
    DOIs
    Publication statusPublished - Mar 3 2014

    ASJC Scopus subject areas

    • General

    Fingerprint

    Dive into the research topics of 'A comparative analysis of computational approaches and algorithms for protein subcomplex identification'. Together they form a unique fingerprint.

    Cite this