Hi,
I'm interested gene ranked list comparison, these lists come from a DEG analysis (log2 ordered).
Normally I would select DEG as gene sets (up or down) from one list and through GSEA I would test for enrichment against a the other pre-ranked list.
In this case I would like to compare two full ordered (log2) lists without selecting genes. Extremes of the list must be weighted, is more important that extremes of the list look alike than the middle of the rank (which should be accounted but less than the extremes).
I'm interested in knowing if both list are the same +1, don't look alike 0, or are exactly the opposite -1.
The methods used for rank comparison normally are: Spearman footrule distance and kendall's/kemeny distance, both weighted and unweighted. If I'm not wrong these methods give distances in the range of [0, 1] or [0, infinite] and won't account for cases when the rank is the same but inverted (wont account for the sign of the log2).
Because of this, I've thought in using rank correlation coefficients whose range are [-1,1], which seems what the function weightedCorr of the wCorr packages doest (it accounts somehow for the extremes weighting them).
Can any of you give a hint on this (packages, methods, etc)? Thanks!
