RT Journal Article
SR Electronic
T1 The Gene Expression Deconvolution Interactive Tool (GEDIT): Accurate Cell Type Quantification from Gene Expression Data
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 728493
DO 10.1101/728493
A1 Brian Nadel
A1 David Lopez
A1 Dennis J. Montoya
A1 Hannah Waddel
A1 Misha M. Khan
A1 Matteo Pellegrini
YR 2019
UL http://biorxiv.org/content/early/2019/08/13/728493.abstract
AB The cell type composition of heterogeneous tissue samples can be a critical variable in both clinical and laboratory settings. However, current experimental methods of cell type quantification (e.g. cell flow cytometry) are costly, time consuming and can introduce bias. Computational approaches that infer cell type abundance from expression data offer an alternate solution. While these methods have gained popularity, most are limited to predicting hematopoietic cell types and do not produce accurate predictions for stromal cell types. Many are also limited to particular platforms, whether RNA-Seq or specific microarray models. To overcome these limitations, we present the Gene Expression Deconvolution Interactive Tool, or GEDIT. Using simulated and experimental data, we demonstrate that GEDIT produces accurate results for both stromal and hematopoietic cell types. Moreover, GEDIT is capable of producing inputs using RNA-Seq data, microarray data, or a combination of the two. Finally, we provide reference data from 7 sources spanning a wide variety of stromal and hematopoietic types. GEDIT also accepts user submitted reference data, thus allowing deconvolution of any cell type, provided that accurate reference data is available.Author Summary The Gene Expression Deconvolution Interactive Tool (GEDIT) is a software tool that uses gene expression data to estimate cell type abundances. The tool accepts expression data collected from blood or tissue samples and sequenced using either RNA-Seq or microarray technology. GEDIT also requires reference data describing the expression profile of purified cell types. Several reference matrices are provided with this publication and on the tool’s website (webtools.mcdb.ucla.edu), and the user also has the option to supply their own. The tool then applies a linear regression to predict which cell types are present in the tissue sample, and in what proportions. GEDIT applies several novel techniques and outperforms other tools on test data.