
Check paper for which yeast samples were "bad" and be sure not to use them

Kallisto abundances.tsv

Test script (Download SRA data?)

P-values
    Paul:
	EdgeR, etc. very sophisticated
	Rank sum test almost as good
	Could be especially useful for large studies with many samples
	where R packages become too cumbersome
	R function wilcox.test()
	    Some cases not as sensitive
	    Some cases more robust
	https://genomebiology.biomedcentral.com/articles/10.1186/s13059-022-02648-4
	Saves as DE-false-positives.pdf
    Paper:
	Wilcoxon test is the only method that controls FDR consistently
	Wilcoxon has almost no power for sample sizes < 8
	    Need a parametric test?
	    Better power than most other methods for SS > 8

Better abundance estimates from BAMs
    Kallisto reports TPM
    https://haroldpimentel.wordpress.com/2014/05/08/what-the-fpkm-a-review-rna-seq-expression-units/
    Effective length l~[i] = l[i] - ufl + 1
	Accounts for all possible start sites of fragments
	ufl = mean fragment length
	bias?
    TPM = transcripts per million
	Adjusts for read count in the sample
	Rate (counts per base) = X[i]/l~[i]
	TPM = rate * (1 / sum(rates)) * 10^6
    FPKM = fragments per kilobase of EXON per million
	= X[i]/(l~[i]*N) * 10^9
	    N = total reads
	RPKM == FPMK for single-end, not meaningful for paired-end
    These are not comparable across experiments due to potential
    differences in RNA distributions under different experimental conditions.
    
