NEW YORK (GenomeWeb) – Using a computational approach they developed, researchers from the University of Pittsburgh have uncovered more than 500 protein-protein interactions linked to genes associated with schizophrenia.
As they reported in NPJ Schizophrenia today, Pittsburgh's Madhavi Ganapathiraju and her colleagues examined the interactomes of genes that had been linked to schizophrenia through either genome-wide association studies or earlier-stage studies to uncover biological pathways that might involved in the disease as well as highlight possible treatment targets.
"GWAS studies and other research efforts have shown us what genes might be relevant in schizophrenia," Ganapathiraju, an assistant professor of biomedical informatics, said in a statement. "What we have done is the next step. We are trying to understand how these genes relate to each other, which could show us the biological pathways that are important in the disease."
She and her colleagues developed a computational model they dubbed High‐Precision Protein Interaction Prediction, or HiPPIP, that characterizes features of a certain protein pair and then relies on a random forest model to gauge whether that pair interacts or not. When they evaluated HiPPIP, the researchers found it to have a precision of 97.5 percent and a recall of 5 percent.
They applied HiPPIP to 108 loci recently linked to schizophrenia by the SZ Working Group of the Psychiatric Genomics Consortium as well as to 25 genes that have been historically linked to the disease. These gene sets have one overlapping gene, the researchers noted.
Through this, they predicted 504 new protein-protein interactions, many of which involved genes that hadn't had any known protein-protein interactions. This, the authors noted, adds to the 1,400 or so known protein-protein interactions in schizophrenia.
The researchers experimentally validated nine of the interactions they uncovered through co-immunoprecipitation or co-localization studies.
Ganapathiraju and her colleagues also examined the gene ontology (GO) terms enriched among the interacting partners of each gene.
"We can infer what the protein might do by checking out the company it keeps," she said. "For example, if I know you have many friends who play hockey, it could mean that you are involved in hockey, too. Similarly, if we see that an unknown protein interacts with multiple proteins involved in neural signaling, for example, there is a high likelihood that the unknown entity also is involved in the same."
For instance, she and her colleagues found that the genes PHOSPH9 and PRRG2 — which have no known GO terms — have interacting partners whose GO terms are enriched for the regulation of ion channels, sodium ion transport, and negative regulation of voltage-gated calcium channel activity, among others.
Overall, they noted that the schizophrenia interactome includes pathways involved in neuronal function and development, dopamine signaling, and immune system and inflammation-related pathways.
Though studying all the previously identified protein-protein interactions reveals all the significant biological pathways the researchers uncovered, folding in the novel interactions uncovers a greater degree of connectedness. For instance, the addition of the newly uncovered protein-protein interactions reveals eight more connections between schizophrenia-linked genes and genes in the RAR activation pathway.
In addition, though the GWAS and historical gene sets only had one overlapping gene, the researchers found that there is a high degree of overlap between their individual interactomes, with 109 genes in common.
A number of genes in these interactomes could represent drug targets, the researchers noted. The GWAS interactome, they found, includes 74 genes that are targets of 307 unique drugs, including nervous system and other drugs. Then by examining what interacts with those drug targets, they identified additional potential drug targets, including NAB2, HRH1, and SCN1A.