how to compare two gene lfc rank is similar or opposite ?

Question

[shangguandong1996]

Hi, everyone

If I have a sorted gene lfc rank from differential gene expression analysis A VS B, like

Gene lfc

Gene1 9.7

Gene2 9.4

……

Gene3 -9.8

And I have another sorted gene lfc rank from C VS D, like Gene lfc

Gene3 4.8

Gene1 3.8

……

Gene2 -2.1

And how can I decide whether two rank set is similar tendency or opposite tendency ?

Best wishes :)

Guandong Shang

Answer 1

[thyleal]

You want to compare gene's modulation between two comparisons, right? Your lists look ordered, but they do not look like ranks (expecting integers). If you want work with ranks, see RankProd or RankSum approaches. Also, check these:

https://www.bioconductor.org/packages/release/bioc/vignettes/OrderedList/inst/doc/tr_2006_01.pdf

https://academic.oup.com/nar/article/38/17/e169/1033168

If you can use the original logFC continuous values, this is what I could think of:

• lollipop/bar plot of genes (y) and logFC(x). Genes with same modulation will have lollipops/bar in the same side of the plot:
• Correlation coefficient of the two gene lists;
• Lin's Concordance Coefficient of the two lists (recommended over correlation because it checks intercepts and slopes);
• Two column heat plot/heat map with a divergent color palette (will help spot visually the discrepancies);
• Scatter plot of the two comparisons and color points by same signal or not (logFC + or -);
• Meta-analysis using effect sizes? I think RankProd and REM/FEM meta-analysis methods are effect size based and will penalize discordant effect signs (not sure, only two comparisosn and come from same dataset, orthogonality etc).

Hope I can at least give you some starting ideas.