However, one shortcoming of the FPKM approach is that if a few very highly expressed transcripts are differentially expressed across samples, the expression level of other transcripts as represented by the FPKM value can be distorted leading to false-positives and false-negatives.
To identify the differentially expressed TFs between different samples, the expression level of all TFs were homogenized by using their RPKM values.
To generate a heatmap plot on three or more samples, the expression values were normalized per gene over all samples.
We found that in HCC and cirrhotic samples, the expression of PARP14 was significantly higher than that in normal quiescent liver samples (GSE6764 34 (Fig. 1a).
Although we know that miR-122 is highly abundant in bulk liver samples, the expression of miR-122 may vary between hepatocytes and this would give some mutants a replicative advantage.
In clinical samples, the expression of EZH2 and Notch1 are positively correlated.
In the double-infected shrimp samples, the expression of HSP-70 in the gills is induced by 3 orders of magnitude when IHHNV is the predominant epizootic, which reveals a relatively strong positive correlation (p < 0.0001 and r = 0.7370) between the number of IHHNV particles and the increase in HSP-70 copy number (Figure 3A).
For most samples the expression levels of lymphocyte markers were equal.
To correct for differences in both RNA quality and quantity between samples, the expression levels of transcripts were normalised to the reference GAPDH gene transcripts.
In fact, in 3 out of 40 samples the expression of CYP3A43 was several fold higher than that of CYP3A4 (Fig 4a).
A common technique in RT-PCR, northern blots or western blots is to normalize data for such artifacts by measuring in the same samples the expression of a reference gene in parallel.
