Re-annotation is an essential step in systems biology modeling of functional genomics data

Abstract

One motivation of systems biology research is to understand gene functions and interactions from functional genomics data such as that derived from microarrays. Up-to-date structural and functional annotations of genes are an essential foundation of systems biology modeling. We propose that the first essential step in any systems biology modeling of functional genomics data, especially for species with recently sequenced genomes, is gene structural and functional re-annotation. To demonstrate the impact of such re-annotation, we structurally and functionally re-annotated a microarray developed, and previously used, as a tool for disease research. We quantified the impact of this re-annotation on the array based on the total numbers of structural- and functional-annotations, the Gene Annotation Quality (GAQ) score, and canonical pathway coverage. We next quantified the impact of re-annotation on systems biology modeling using a previously published experiment that used this microarray. We show that re-annotation improves the quantity and quality of structural- and functional-annotations, allows a more comprehensive Gene Ontology based modeling, and improves pathway coverage for both the whole array and a differentially expressed mRNA subset. Our results also demonstrate that re-annotation can result in a different knowledge outcome derived from previous published research findings. We propose that, because of this, re-annotation should be considered to be an essential first step for deriving value from functional genomics data.

Figure 1. Whole microarray GOSlim modelling.

The difference in number of GO annotations in the GOSlim groups for the GO ontologies ‘cellular component’, ‘molecular function’ and ‘biological process’ between the original and re-annotated whole microarray gene dataset. The whole microarray GOSlim modeling shows that re-annotation increases the number of GO annotations in each GOSlim group for each ontology.

Figure 2. Differentially Expressed mRNA GOSlim modelling.

The difference in number of GO annotations in the GOSlim groups for the GO ontologies ‘cellular component’, ‘molecular function’ and ‘biological process’ between the original and re-annotated differentially expressed mRNA dataset. The differentially expressed mRNA GOSlim modeling shows that re-annotation increases the number of GO annotations for most GOSlim group. The negative value for the GOSlim group ‘transporter activity’ in the ‘molecular function’ ontology are caused by updated GO annotations to the more detailed ‘protein transporter activity’ GOSlim group.