Fig. 1
From: Annotation depth confounds direct comparison of gene expression across species
Transcripts per million (tpm) values for 10 sample genes appear inflated in sub-sampled transcriptomes, compared to their estimate in full length human transcriptome. A Gene level estimates of log10(tpm) difference for the 10 selected genes (individual lines and colors) from their estimate in the fully described human transcriptome. We selected genes through stratified random sampling from 10 bins of final abundance estimate across 9900 iterations of transcriptome with sizes from 1–99% (number of transcripts) of original transcriptome size. B Log10 difference in number of reads mapped to those genes in the same sampling scheme reveals that the large shifts in tpm estimates are driven by shifts in number of reads mapped. However, subtle and consistent shifts are independent of number of reads mapping to those genes