Neuroinformatics for gene expression: networks, function and meta-analysis

University Of British Columbia, Vancouver BC

Investigators

Linked publications & trials

Abstract

Summary This project helps address challenges neurobiologists face when trying to interpret the results of genomics and genetics studies, in which a typical aim is to identify candidate genes and regulatory pathways relevant to normal or disease processes. A common approach is to use gene network analysis to position findings in a systems biology context. However, identification of relevant patterns in complex data remains difficult. Our past work resulted in a large database of expression profiling data, ?Gemma? (now over 10,000 studies, including 3,200 relating to the nervous system) and an extensive series of analyses. This work led to the identification of specific biases and confounds affecting gene network analysis and gene annotations, and the development of improved approaches and tools for gene function prediction, gene set enrichment and cellular composition analysis. Recently we have shown that much of the variance of expression in brain data sets comes from differences in cellular composition among samples. Building on this and other work, we propose to extend our research to further the interpretability of gene expression patterns in the nervous system. Our first aim is to continue the development of our transcriptomic re-analysis databases, with the expected addition of approximately 2,500 additional brain-related studies, and an expansion into single-cell transcriptomics. Our second aim is to assemble and evaluate a high-quality dataset of transcription factor regulatory targets in brain using a combination of manual curation of the literature and integration of chromatin immunoprecipitation with transcription factor perturbation studies along with third-party sources. The third aim brings these topics together to further the extent to which transcriptomic data can be used to infer changes in regulatory influences in both single-cell and bulk-tissue data. We hypothesize that our approaches will allow increased resolution of transcription-factor regulated patterns of co-expression and differential expression. We will then apply these interpretational tools to re-analyze bulk and single-cell data sets in neuropsychiatric and neurodevelopmental conditions, and to study candidate genes identified via genetic association of rare or common variants. All of our data and software will be freely disseminated via web applications, programmatic interfaces and downloadable data sets.