Fig. 4From: Improving sensitivity of linear regression-based cell type-specific differential expression deconvolution with per-gene vs. global significance thresholdDropping cell proportions reduces condition number and improves sensitivity. Effect of dropping (a) 1 cell type; (b) 3 cell proportions with the lowest proportion mean. The following parameters were used: log2-transformed data, per-group sample size – 10, MSE – 1.5, cell type-specific effect ranged from 0.001 to 1.0 over 500 genes out of 1000 simulated genes, SD of the cell type with introduced changes – 0.1, condition number (5 cells total) – 75000Back to article page