Bioinformatics Seminar
Time:
Venue: Na
4 October 2016
NaEstablishing Cell Composition Via DeconvolutionĀ
Saskia FreytagWEHI Population Health & Immunity Division
Both microarray and RNA-seq technology are frequently used to measure gene expression in samples consisting of many different cell types. However ; gene expression differs between cell types to create the specialized functions and physiologies of many different cell types. Differences in gene expression due to cell type have been demonstrated to drive differential expression results in a recent study of psychiatric disorders (Bowling et al ; bioRvix 2016). Given that cell composition can also change as a result of disease ; e.g. schizophrenia (Joshi et al ; Biological Psychiatry 2012) ; it is important to account for differences in cell-type proportions in differential expression analysis. Since sample compositions are rarely measured ; they have to be accurately inferred using computational algorithms ; referred to as deconvolution.
We recently developed a new simple deconvolution approach based on ratios. We demonstrate that this new approach is able to infer cell composition more robustly than any of the existing approaches in a variety of test datasets with true cell proportions established either by fluorescence-activated cell sorting or by experimental design. Moreover ; our approach is fast ; scalable and easy to optimize. Finally ; we show that it can be applied to a large gene expression dataset obtained from post-mortem schizophrenic and normal brain (Fromer et al ; bioRvix 2016) and change results on differentially expressed genes.