Some genes are known to be key players in complex diseases like glioblastoma multiforme — they have roles in one or more pathways known to be dysregulated. But which among these genes, if any, are master regulators?
That's what the National Center for Biotechnology Information's Teresa Przytycka and her colleagues are trying to learn. In PLoS Computational Biology in March, Przytycka's team presented its pathway-centric computational approach — based on expression quantitative trait loci analysis, graph theoretical techniques, and combinatorial algorithms — to identify causative genes in complex diseases.
At the National Institutes of Health's intramural Research Festival in October, Przytycka outlined her group's methodology. Rather than single genes, she and her colleagues interrogate pathways as markers of disease. For any given complex disease, the team searches for starting points in the literature. "What we find most useful is to start with selecting a set of target or disease-marker genes, because that helps you have higher statistical power," Przytycka said. "We start with those genes that we selected, but then add the pathways from genetic mutations to selected genes, because in every patient that may be different. Since we now have a hold of the proposed individual pathways, we use those as markers to see how they overlap."
Still, there are a number of genes critical for disease pathway signaling that are not themselves up- or down-regulated and that gene expression studies therefore do not pick up. These genes, Przytycka said, could act as "mediators of signal flow to target genes." For that, the team looks to genes that overlap several pathways, or what Przytycka called hubs. She added that "once we have those sets, the most reasonable thing to do would be to check out the pathway annotation." That's where her group's computational method comes in.
In their paper, the researchers reported that their approach, when applied to genomic alteration and gene expression profiles for 158 glioblastoma multiforme patients, allowed them to identify not only putative causal genes, but also potential "intermediate nodes and pathways mediating the information flow between causal and target genes."
At NIH, Przytycka added that applying a pathway-centric approach takes researchers a step beyond gene expression analyses and may help them identify master regulators. "If a gene has lots of neighbors that are correlated, but itself is not, it's still very likely to be used in the propagation of the information," Przytycka said. And just because a gene is not an expression analysis hit, that should not discount its potential importance, she added.