Ontology in genomic and proteomic experiments: Which reference list?

FRESNO C; LLERA AS; GIROTTI MR; PODHAJCER O; BALZARINI MG; PRADA F; FERNÁNDEZ EA

Búzios, Rio de Janeiro, Brasil

Congreso; Brazilian Symposium on Bioinformatics (BSB'10) - International Workshop on Genomic Databases (IWGD'10) - Escola Brasileira de Bioinformática (EBB'10); 2010

Ontology analysis is currently one of the main steps in most of the genomic/proteomic experiments. It is usually carried out to relate identified differentially expressed genes to their functional counterparts, identifying enriched functions, pathways, etc. by the experiment. The required data for statistical validation of such functional information is acquired by querying large databases with controlled vocabulary such as Gene Ontology (GO) [1], where genes and their known functionality, location, etc. are stored. There are plenty of bioinformatic tools to deal with them such as DAVID,GOMINER, etc[2]. These tools entail the fulfillment of different requirements, being one of them the selection of the Reference Gene List (RGL), used to contrast the Differentially Expressed Gene List (DEGL). The RGL is used as the reference ("universe") to compare the proportions of genes in the represented categories between the DEGL and the "universe", expecting that some of them get enriched in the DEGL in contrast to the RGL situation. This suggests that the results could vary according to the chosen RGL. The appropriateness of the RGL to determine reliable enriched ontologies has not been acknowledged yet. Different RGL choices can be made according to the experiment carried out. In the case of microarrays studies the possibilities are: (i) use the whole genome reference gene list (GRGL ) from the corresponding organism (the default option in many tools), (ii) use the chip reference gene list (CRGL) or (iii) use only those genes reliably detected (RDGL ) in the experiment. In the case of high-throughput proteomic studies such as gel based protein expression, the CRGL is not available. In some experiments (i.e. secretome studies), only a small subset of the genome is evaluated, so, the selection of the appropriate RGL is not clear in this case. In this work we compare some of these situations to evaluate their impact on ontology analysis, through a microarray experiment designed to study wound healing in D. melanogaster.