search

hms-dbmi / scde

R package for analyzing single-cell RNA-seq data
http://pklab.med.harvard.edu/scde
Other
170 stars 64 forks source link

Significant corrected Z-score? #50

Closed livnatje closed 6 years ago

livnatje commented 6 years ago

Thank you for providing this great pipeline!

Could you recommend which cutoff to use? What would be considered a significant corrected Z-score?

JEFworks commented 6 years ago

Of course. Thanks for using it!

I would generally recommend a cutoff of 1.96 corresponding to a p-value of 0.05 for a two-tailed test.

Here is a good visualization of how p-values relate to z-scores and relate to normal distributions: image (http://pro.arcgis.com/en/pro-app/tool-reference/spatial-statistics/what-is-a-z-score-what-is-a-p-value.htm)

If you are looking for strong markers, a z-score of 3 may be more appropriate. If you are doing gene set enrichment afterwards or looking for general trends among your differentially expressed genes, a z-score of 1.28 may be more appropriate.

livnatje commented 6 years ago

Thank you Jean! So, just to make sure, cZ would be after multiple hypotheses correction, so a cutoff of 1.96 would make sense, right?

JEFworks commented 6 years ago

Yes exactly! You will likely want to consider both below -1.96 (significantly downregulated) and above 1.96 (significantly upregulated) ie. absolute cZ > 1.96