Sparse representation based biomarker selection for schizophrenia with integrated analysis of fMRI and SNP data

We propose a novel sparse representation based variable selection algorithm (SRVS), which improves the variable selection ability of a traditional sparse regression model in that it performs variable selection at different significance levels, and gives groups of selected variables of different sizes. As an example, we applied the algorithm to a joint analysis of 759075 SNPs and 153594 functional magnetic resonance imaging (fMRI) voxels in 208 subjects (92 cases/116 controls) to identify biomarkers for schizophrenia (SZ). To evaluate the selected biomarkers, a 10-fold cross validation was performed. The results between SRVS method and a previously reported variable selection method were compared, which showed that our method, especially with a sparse regression model penalized with norm, gave significantly higher classification accuracy of discriminating SZ patients from healthy controls.