Investigating the genetic interactions of schizophrenia using gene-based statistical epistasis

The interplay between genotype and phenotype is governed by a multitude of genetic interactions (GIs), and the mapping of GI networks holds significant importance for two main reasons: (1) GIs offer a valuable means to uncover compensatory biological mechanisms by modelling biological robustness, thereby identifying functional relationships between genes. This aspect is particularly relevant for biological exploration and translational research, as biological systems have evolved to compensate for genetic (i.e. variations, mutations) and environmental (i.e. drug efficacy) perturbations by leveraging compensatory relationships between genes, pathways, and biological processes; (2) GI facilitates the identification of the direction (positive/alleviating or negative/aggravating interactions) and magnitude of epistatic interactions that influence the resulting phenotype. While comprehensive GI databases exist for organisms like yeast, generating GIs for human diseases through experimental biology methods such as systematic deletion analysis is infeasible. Furthermore, generating disease-specific GIs in humans has not been previously attempted. We used the Indian schizophrenia case-control (case-816, controls-900) genetic dataset to implement and test GI workflow. Standard GWAS sample quality control procedure was followed to check for ancestry and relatedness outliers. We used the imputed genetic data to increase the SNP coverage to analyse epistatic interactions across the genome comprehensively. By using the odds ratio (OR) we identified the GIs that increase (OR >1) or decrease (OR < 1) the risk of a disease phenotype (i.e. schizophrenia). The SNP-based epistatic results were transformed into gene-based epistatic results. We have developed a GI workflow for conducting gene-based statistical epistatic analysis and transforming these results to infer GIs. Spatial analysis of functional enrichment (SAFE) was used to detect the statistically overrepresented functional groups. There were ∼ 9.5 million GIs with a p-value ≤ 1 × 10-4. Approximately 4.8 million GIs showed an increased risk (Odds Ratio > 1.0), while ∼ 4.75 million GIs had decreased/no risk (Odds Ratio < 1.0) for schizophrenia. We identified many hub genes with numerous GIs, which increased and reduced the risk of Schizophrenia. In contrast to model organisms, this approach is specifically viable in humans due to the availability of abundant disease-specific genome-wide genotype datasets. Despite limited power, meaningful GI data was generated with a small sample. SAFE and REVIGO analysis exclusively identified brain/nervous system-related processes, affirming the findings. This computational approach fills a critical gap by generating practically non-existent heritable disease-specific human GIs from human genetic data. These novel datasets can train innovative deep-learning models for post-GWAS functional characterisation, potentially surpassing the limitations of conventional GWAS.