2.7 Differential gene expression analysis
Differences in gene expression among groups were identified by
differential expression analysis performed using DESeq2 on raw read
counts (non-normalized, as suggested by DESeq2). To allow comparison
between QuantSeq reads and NEB reads data, we only used one sequencing
direction as suggested for the DESeq2 program for the latter. The false
discovery rate (FDR) was adjusted to 0.05, corresponding to a recovery
at most of 5% of false positives following the DESeq2 manual. We use
the default options for all the other parameters. We look at differences
in gene expression between sampling methods, harvest tissue time, tissue
type, and QuantSeq vs. NEB in Table 2: (see Tables 1, 2 and data on
Dryad per detailed information about comparisons and sample size for
each comparison; minimum N = 4). The log2 fold changes obtained from
DESeq2 were used as a measure of how many more (or less) genes are
expressed in one group versus the other. We considered genes having
different expression if the adjusted p-value (using the adjusted p-value
results in less false positives) was < 0.05.
Finally, previous work has indicated an increase in read count for
longer transcripts using NEB than QuantSeq (Ma et al. 2019; but see Crow
et al., 2022). To further address the relationship between gene length
and genes differentially expressed between QuantSeq and NEB, we
conducted an assessment using the known transcript length from
orthologous genes in zebrafish in Ensembl 101 (Yates et al. 2020) based
on gene name for genes that were detected to be differentially expressed
between the two library types. We used the zebrafish instead of the
rainbow trout genome as the former has a more curated annotation (and
thus more precise gene length information) than the latter. We also used
the same approach to specifically assess if transcript length could
influence absence of gene expression or not detection (mean base = 0 in
DESeq2 output) in one but not the other library type, QuantSeq or NEB.