Multilocus Analysis of Case-control Data to Schizophrenia Based on Measures of Information Discrepancy

As one of complex diseases, schizophrenia is considered to possess a complex trait to which genetic, environmental, and epigenetic factors contribute interactively. The genetic analysis of schizophrenia has revealed complex and inconsistent results, making it difficult to draw clear conclusions regarding the impact of specific genes on the disease in diverse human populations. So the need for identification of susceptibility genes especially multilocus interactions for schizophrenia still poses a great and unanswered challenge. In this paper we propose a new approach to characterize the individual and combined contributions of three recently proposed candidate genes (NRG1, G72 and RGS4) in the schizophrenia case-control dataset in the Chinese Han population. The approach is based on measures of information discrepancy (MID). First the susceptible loci are selected based on MID by taking into consideration the gene–gene interactions. Then a discrimination algorithm is introduced to assess the effectiveness of the selected multiple loci to predict the disease status, that is, a classification accuracy is obtained by reclassifying the subjects into the case or control groups. Several possible pathogenic locus-locus interactions for schizophrenia are obtained.