Biomarker discovery in heterogeneous tissue samples -taking the in-silico deconfounding approach

BMC Bioinformatics

Table 1 Profile reconstruction versus differential gene expression: alternatives for deconfounding algorithm settings

	Optimal deconfounding algorithm settings
	log/quant	log/not quant	not log/quant	not log/not quant
cor reconstr	0.86	0.85	0.73	0.68
DGE power	0.47	0.46	70	0.62

Correlations of measured and estimated cell type-specific gene expression profiles ("cor reconstr.") as well as power for detection of differential expression ("DGE power", see text) -- for all four combinations of using logs or not, applying quantile or global mean normalization, respectively, for the deconfounding algorithm.

ISSN: 1471-2105