Integrative Analysis of DiseaseLand Omics Database for Disease Signatures and Treatments: A Bipolar Case Study

We presented here an integrative analysis framework of the Janssen DiseaseLand transcriptomics database, which currently includes 1,885 human and 1,460 animal (mouse or rat) studies from public and Janssen internal sources, and is constantly growing. By applying standardized re-processing to raw datasets, removing outlying samples and problematic datasets, and comparing multiple meta-analysis approaches, we were able to generate a unified gene expression signature for a disease. We demonstrated our approach on a meta-analysis of 30 gene expression datasets from human brain tissues in bipolar. Bipolar is a serious mental illness with considerable public health implications. However, our understanding of biological mechanisms of bipolar remain frustratingly limited in part due to difficulty in accessing human brain samples (Wu et al., 2017a). Many individual bipolar transcriptomics studies contain only tens of samples (Table 1), which may contribute to a lack of reproducibility in genes and pathways identified from each study. By combining these smaller studies through meta-analysis, we can additionally reduce the effects of heterogeneity in platforms and pipelines among studies. By applying our proposed meta-analysis framework to a total of 30 existing studies of bipolar in the human brain, we generated a highly reliable gene signature that is associated with bipolar. While our studies encompassed tissue from several brain regions, we saw little pathway enrichment in regions...
Source: Frontiers in Genetics - Category: Genetics & Stem Cells Source Type: research