How to define doublets after getting the score?

kostkalab / scds

In-silico doublet annotation for single cell RNA sequencing data

Other

8 stars 2 forks source link

How to define doublets after getting the score? #2

Open YiweiNiu opened 4 years ago

YiweiNiu commented 4 years ago

Hi,

Thank you for developing this useful tool!

A quick question: how to define doublets after getting the score? Should we use something like normal distribution to exclude the outliers? I want to hear your advice.

I am new to this. Sorry if the question is too obvious.

Bests, Yiwei Niu

GildasLepennetier commented 3 years ago

Hi there! I am also a bit confused. I know I have doublets in my data (19007 features across 20827 samples), but the histogram of the scores (CD$cxds_score,CD$bcds_score, CD$hybrid_score) display a large peak at , let say, a cxds_score~=17000, with scores rising up to 27000. Do we want then to select all cell above a score threshold (and e.g. representing 7% of total cell)?

zhangguy commented 3 years ago

Hi,

Thank you for developing this useful tool!

A quick question: how to define doublets after getting the score? Should we use something like normal distribution to exclude the outliers? I want to hear your advice.

I am new to this. Sorry if the question is too obvious.

Bests, Yiwei Niu

Also want to know how to determine a cutoff for the score, especially when there are many samples.

kenneditodd commented 2 years ago

@kostkalab @nturaga @asbDBPub Can you answer the above questions?

kostkalab commented 2 years ago

Apologies, just got the notification (see above). Not sure if there is still interest, but if so let me know. Here is a short answer:

The is certainly more than one reasonable way to choose a cutoff.
One that we have implemented is to use cxds (or bcds, or cxds_bcds_hybrid) with the parameter estNdbl = TRUE.
This will generate binary doublet calls in addition to the doublet scores.