Fig. 2

Bioinformatics QC to identify pass versus fail samples. a Heatmap of sample pairwise correlation of gene expression. Row color annotation bar indicate sequencing batch (seqb) 1–7 and 8–10. Right lower panel shows a histogram of the distribution of sample wise median correlation based on gene expression data. b Relationship between sample wise median correlation of gene expression with false positive rate using 11 study replicate samples. Samples with a sample-wise median correlation below 0.75 were classified as QC failed samples. Loess is used curve fitting and 95% confidence interval is plotted in grey bands. c Relationship between sample wise median correlation of gene expression with number of detectable genes with transcript per million (TPM) > 4. A cutoff of 11,400# of genes was selected to identify QC failed samples. d Relationship between number of gene mapped reads and total number of detected genes with transcript per million (TPM) > 4. A cutoff was selected at 80% of saturation point (20 million gene mapped reads, 10,400 # of detected genes with TPM > 4)