Add QC to code and tutorial

[oganm]

• Low expressed genes warning and documentation
• PC variance explained warning and elaborate the QC steps in tutorial
• Gene removal variable that defaults to genes found not to be specific in humans. (justify with either external data (Darmanis) or gene robustness tests)
• Add subsampling of samples to the tutorial.

[oganm]

Quantify change in terms of expression levels

Aim is to be able to compare effect sizes

• percent change
• absolute change

Output mean expression values of genes used for estimation.
Output trimmed expression matrices? This might be too much clutter and hard to interpret.
Percent change can confuse people who want to interpret is at percent change in cell counts.

However it is difficult to compare effect sizes between two studies as they would be using different sets of genes. So average expression of markers would not be comparable.

Trimmed expression matrices could be useful to pick genes that are used in studies, while mean expression can be a more general measure of effect size