Measuring the wisdom of the crowds in network-based gene function inference

• DOI: 10.1093/bioinformatics/btu715
• Publication Date: 2014-10-31T00:44:56Z
• Authors: W. Verleyen; S. Ballouz; J. Gillis
• Keywords: Code (set theory)

Abstract Motivation: Network-based gene function inference methods have proliferated in recent years, but measurable progress remains elusive. We wished to better explore performance trends by controlling data and algorithm implementation, with a particular focus on the of aggregate predictions. Results: Hypothesizing that popular would perform well without hand-tuning, we used well-characterized algorithms produce verifiably ‘untweaked’ results. find most state-of-the-art machine learning obtain ‘gold standard’ as measured critical assessments defined tasks. Across broad range tests, see close alignment performances after for underlying being used. aggregation provides only modest benefits, 17% increase area under ROC (AUROC) above mean AUROC. In contrast, gains are enormous an 88% improvement Altogether, substantial evidence support view additional development has little offer prediction. Availability implementation: The supplementary information contains description algorithms, network parsed from different biological resources guide source code (available at: http://gillislab.cshl.edu/supplements/). Contact: [e-mail]