Automated GO annotation based on graphical models

SPETALE FLAVIO EZEQUIEL; KRSTICEVIC FLAVIA; ANGELONE LAURA; BULACIO PILAR; TAPIA ELIZABETH

Rivera Maya

Conferencia; Bioinformatics & Computational Biology for Innovative Genomics; 2016

Automated annotation of coding genes based on Gene Ontology (GO) assigns biological functions throughout bioinformatic methods. Standard methods for automated annotation of coding genes commonly rely on sequence similarity or protein signature searches. However, in absence of clear sequence similarities or definite protein signatures, alternative computational methods for automated gene annotation must be considered. These difficulties may be circumvented by using hierarchical ensemble classification techniques. By means of these techniques, the problem of automated gene annotation can be cast to that of predicting individual GO-terms within a True Path Graph (TPG), a special kind of Directed Acyclic Graph (DAG) defining the meaning of GO-terms by multiple inheritance. Predictions of individual GO-terms are expected to be noisy and inconsistent with the underlying TPG and need to be leveraged. A Factor Graph approach to the automated GO Annotation(FGGA) of protein-coding genes has been proposed.Evaluations on three genomes model organisms of protein sequences from the GO Molecular Function domain show significant improvements