Additional file 8: of CellMapper: rapid and accurate inference of gene expression in difficult-to-isolate cell types
Year: 2016
Type: article
Abstract: Robustness of CellMapper to dataset size and composition. CellMapper was applied to predict genes expressed in each of the eight brain cell types, using random subsets of (left) the Allen Brain Atlas dataset or (right) an independent expression compendium of 1237 human brain samples drawn from the Lukk et al. [28] and Engreitz et al. [27] datasets (the “Affymetrix Brain Samples” compendium). The accuracy with which each search identified the experimentally-defined cell type genes in mice [3–6] was then quantified by the area under the precision-recall curve (AUPR). AUPR was calculated for 50 randomly sampled datasets of the indicated sample sizes and then mean AUPR was calculated. Results are reported as a heatmap, with all AUPRs scaled relative to the performance achieved when using the complete Allen Brain Atlas dataset for each cell type. The sensitivity to dataset abundance varied, with maximum AUPR being reached between 100 and 2000 microarray samples depending on the cell type. Overall, we conclude that a high quality, large, and uniformly collected dataset such as the Allen Brain Atlas is likely to allow for accurate predictions for a wider range of cell types. Black squares indicate that the AUPR was not significantly different from chance (Bonferroni corrected p value > 0.05; permutation test). (PDF 81 kb)
Source: Figshare
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7137841009