PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Weighted True Path Rule: a multilabel hierarchical algorithm for gene function prediction
Giorgio Valentini and Matteo Re
In: MLD-ECML 2009, 1st International Workshop on learning from Multi-Label Data, 7 Sept 2009, Bled, Slovenia.

Abstract

The genome-wide hierarchical classification of gene functions, using biomolecular data from high-throughput biotechnologies, is one of the central topics in bioinformatics and functional genomics. In this paper we present a multilabel hierarchical algorithm inspired by the “true path rule” that governs both the Gene Ontology and the Functional Catalogue (FunCat). In particular we propose an enhanced version of the True Path Rule (TPR) algorithm, by which we can control the flow of information between the classifiers of the hierarchical ensemble, thus allowing to tune the precision/recall characteristics of the overall hierarchical classification system. Results with the model organism S. cerevisiae show that the proposed method significantly improves on the basic version of the TPR algorithm, as well as on the Hierarchical Top-down and Flat ensembles.

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Theory & Algorithms
ID Code:6536
Deposited By:Giorgio Valentini
Deposited On:08 March 2010