Additional file 2: Table S2. of Improving sensitivity of linear regression-based cell type-specific differential expression deconvolution with per-gene vs. global significance threshold

An example of LRCDE and csSAM analysis results of simulated data with 18 samples per group. Target condition number (kappa) – 100, target mean squared error – 1.5, target cell proportion SD – 0.1. Range of the effect sizes introduced in the first 500 genes – 0.001 to 1.0. “site” – Probe ID; “Gene” – Gene label; “base” – estimated non-negative control group cell type-specific expression level; “case” - estimated non-negative case group cell type-specific expression level; “diff.est” – estimated cell type-specific differential expression between case and base; “mse.control” – Observed mean squared error of the control group regression; “mse.case” – Observed mean squared error of the case group regression; “cell” – Cell type name; “cell.sd” – Cell type-specific standard deviation of cell proportions across samples; “kappa.1/2” – Observed condition number of squared cell proportions matrix for the case/control group, respectively; “t.crit” – Critical t-value against which observed t-statistic is compared; “t.stat” – Observed t-statistic; “p.val.t” – Unadjusted p-value for t-statistic; “se1/2”– Standard error of the control/case group cell type-specific expression (regression coefficient) estimate, respectively; “se.p” – Welch’s combined standard error calculated for t-statistic; “t.power” – Observed power calculated from t-statistic and critical t-value; “FDR” – False discovery rate as reported by csSAM. (XLSX 136 kb)