Part 1. For our initial analysis, we parcellated the whole brain into 116 anatomic regions and 1000 functional networks, then analyzed connectivity within and among brain regions and networks. To test our results, we analyzed the same data using a multivariate classification method, which confirmed most of the connectivity features our new method selected. As expected, both comparisons showed widespread differences in brain connectivity between schizophrenia cases and healthy unrelated controls. More unusually, they also revealed unexpectedly large differences between cases and their healthy siblings. This project will continue in FY2014, when (in accordance with formal IRB approval and detailed collaborative agreement) extramural collaborators are scheduled to provide us with new datasets for analysis. Part 2. Following our initial analytic method development, proof-of-concept simulations, and real data validation tests , we now aim to refine our multi-sequenced mixture model for the analysis of series of data points in order to predict the tipping point when, in our test case, disease symptoms or therapeutic drug effects become evident. In extensive simulation tests, our mixture model has produced encouraging results. We will therefore continue to improve our model by making it more flexible in terms of handling missing data and will test its feasibility using real data, including data sets from the massive STAR*D study of persons diagnosed with depressive disorders.
|Cao, Hongbao; Duan, Junbo; Lin, Dongdong et al. (2014) Sparse representation based biomarker selection for schizophrenia with integrated analysis of fMRI and SNPs. Neuroimage 102 Pt 1:220-8|
|Yuan, Jianmin; Jin, Chunhui; Qin, Hai-De et al. (2013) Replication study confirms link between TSPAN18 mutation and schizophrenia in Han Chinese. PLoS One 8:e58785|