Dataset and normalization

The count table of the cells passing the filtering procedure was processed to keep only the genes with average count larger than 1. The normalization uses a pooling strategy implemented in R function computeSumFactors (L. Lun et al., 2016). The normalized data is in log2 space. To remove the patient specific variance, the normalized table was further centered by patient, so in the centered expression table the cells of each patient had zero mean and variance as before centering. The final expression table had 14127 genes and 4008 cells. TPM data is filtered accordingly.