Brain Volume Prediction from SNP network with Semi-Supervised Regression

Dong-Gi Lee, Hyunwoong Roh, Myungjun Kim, Na-Rae Kim, Sang Joon Son, Changhyung Hong, Hyunjung Shin

Abstract: The gray matter volume of the brain is used as one of the important indicators to evaluate cognitive function. That is, the smaller the gray matter volume in the region of memory, the lower the cognitive function. Although there are several factors in the reduction of gray matter volume, it has been found that genetic factors also play a role through numerous recent studies. Genetic factors can involve in biological activities not only independently, but also collectively with complex interactions. In this study, we propose a method for predicting brain volume and deriving significant genetic variants from single-nucleotide polymorphisms (SNP) network that reflects the interactions between SNPs. The proposed method constructs a linear regression model to predict brain volume using refined SNP features obtained through feature propagation on the SNP network. The prediction model was applied to biobank innovations for chronic cerebrovascular disease with Alzheimer's disease study (BICWALZS) participants in Ajou University Hospital, Korea.