Accumulation of minor alleles and risk prediction in schizophrenia

Abstract

Schizophrenia is a common neuropsychiatric disorder with a lifetime risk of 1%. Accumulation of common polygenic variations has been found to be an important risk factor. Recent studies showed a role for the enrichment of minor alleles (MAs) of SNPs in complex diseases such as Parkinson's disease. Here we similarly studied the role of genome wide MAs in schizophrenia using public datasets. Relative to matched controls, schizophrenia cases showed higher average values in minor allele content (MAC) or the average amount of MAs per subject. By risk prediction analysis based on weighted genetic risk score (wGRS) of MAs, we identified an optimal MA set consisting of 23 238 variants that could be used to predict 3.14% of schizophrenia cases, which is comparable to using 22q11 deletion to detect schizophrenia cases. Pathway enrichment analysis of these SNPs identified 30 pathways with false discovery rate (FDR) <0.02 and of significant P-value, most of which are known to be linked with schizophrenia and other neurological disorders. These results suggest that MAs accumulation may be a risk factor to schizophrenia and provide a method to genetically screen for this disease.

Figure 1

Weighted genetic risk score distribution in cases and controls. Distribution of weighted genetic risk score with total SNPs and LD-independent SNPs of case and control subjects in GAIN and MGS cohort. Controls 1 and cases 1: calculated with total SNPs; controls 2 and cases 2: calculated with LD-independent SNPs with r² threshold of 0.3.

Figure 2

Discriminatory abilities of different wGRS prediction models from external cross-validation analysis. Discriminatory abilities of 130 wGRS prediction models constructed by total SNPs (a,b). Discriminatory abilities of 208 wGRS prediction models constructed by LD-independent SNPs (c,d). AUC (a,c) and TPR (b,d) were calculated using a training dataset (GAIN) and a validation dataset (MGS) to evaluate the discriminatory abilities. *The optimal model with the best performance among models constructed by LD-independent SNPs.

Figure 3

Role of intergenic SNPs in prediction performance. (a) AUC values. (b) TPR values. set1: without intergenic SNPs in optimal wGRS models; set 2: intergenic SNPs in optimal wGRS models; total: all SNPs contains in optimal wGRS models.

Figure 4

Top-10 enriched gene modules of gene ontology analysis with SNPs in optimal weighted Genetic Risk Scores model. Significant enrichment of gene modules was analyzed from WebGestaltR based on categories of biological process, molecular function, and cellular component respectively, more details can be found in Supplementary Table S6-1.