RNA-sequencing Example
Data from GEO: Series GSE144748
"The MiDAC histone deacetylase complex is essential for embryonic development and has a unique multivalent structure" Nature Communications (2020)
Quality Control
Normalised count distribution
Distribution of expression values following normalisation
Overlapping / aligned samples confirm normalisation has corrected for library size differences
PCA plot
Reduces data with many dimensions to principal components for visualisation
Samples from the same condition should cluster together making outliers or batch effects obvious
In this example control_2 and KO_2 are slightly skewing results and may need to be corrected for
Given the nature of these samples this is not unexpected (mouse fibroblasts from 3 different litters)
Sample correlation
Pairwise correlation coefficients between samples
Within-group correlations should be high
Inter-group correlations should be lower
Not always easy to distinguish if there are only moderate changes in gene expression
Cook's distance
Measures effect of each sample in fitted model coefficients for every gene
High Cook's distances across many genes may be outliers
This will effect downstream differential expression results
Dispersion estimates
Relationship between each genes mean expression and its biological variability
DESEQ2 - used for differential gene expression analysis - shrinks estimates towards a fitted trend
A plot that shows final estimates clustering tightly around the trend line with only a moderate number of high variance genes above is generally considered a good fit
