search

Hoohm / dropSeqPipe

A SingleCell RNASeq pre-processing snakemake workflow
Creative Commons Attribution Share Alike 4.0 International
147 stars 47 forks source link

advice on knee plot #25

Closed olechnwin closed 5 years ago

olechnwin commented 6 years ago

Although this is not issue on running dropSeq pipeline, however, since the wiki has explanation for the knee plot, I thought I'll post it here in hope of getting advice.

This is the second run of DropSeq that we've analyzed and both times my knee plots never show a clear bend (see attached). knee_plot.pdf What is your advice for this? Thanks!

Hoohm commented 6 years ago

Hello @olechnwin From my experience the knee plot is just "bad" for true STAMP selection.

For now I just leave this as is and propose people to use other selection methods that rely more on metadata variables see example than "simple" counts statistics.

I can suggest you a good read here regarding true STAMP selection.

0.31 will still not propose a new method but check out the new branches that will come up just after the new release and you'll find some new features for true STAMP selection.

I hope this helps out a bit.

We are still thinking about the best way to implement it but for now, this is the best I can give you.

seb-mueller commented 6 years ago

Further to what Patrick has said, we were in fact discussing something similar in issue #18 (especially towards the end). I've implemented a slightly more informative knee-plot on this branch. If you are keen you are welcome to try it out, I'd be curious on any feedback.

olechnwin commented 6 years ago

Thank you @Hoohm for your response. I'll try the selections that you suggested in stackexchange. @seb-mueller, thanks for letting me know. I'll let you know my feedback after I tried your knee-plot.