Include cells with 2+ guides in low MOI analysis

[argschwind]

Hi,

We use Sceptre to analyze an enhancer targeting TAP-seq screen from our lab. It's a low MOI experiment, so naturally we use the low MOI setting in Sceptre. Even at low MOI there's always a small fraction of cells with 2+ guides, but we have noticed that with the low MOI setting these cells are rigorously filtered out.

Is there an easy way you could recommend to include cells with 2+ guides in the discovery analyses?

We also tried running Sceptre in the high MOI setting but observed model calibration issues.

Thanks in advance for you help!

[timothy-barry]

Hi Andreas,

Good to hear from you. I guess this an issue related to the control group. Would the idea be to use cells that contain a (single) NT gRNA as the control group, while keeping cells that contain 2+ targeting gRNAs?

[timothy-barry]

I see. OK, I will allocate an hour on Wednesday to tinkering around with this. I will let you know what I come up with.

[argschwind]

Thank you so much for looking into this, Tim!

[argschwind]

Hi, are there any updates on this or could you let us know when we can expect some? I understand that you're probably busy with many other important things as well.

[timothy-barry]

Hey Andreas, working on this now. Nothing to share yet.

[timothy-barry]

I put some time into this today, and unfortunately, I think it will be challenging to implement this update to the package. I am going to give this another shot on Friday, but sadly I can't make any promises that this functionality will be available in the near future.

Do you think you might be able to take a crack at implementing this functionality? If so, I can point you to the relevant parts of the code over a Zoom call.

[timothy-barry]

Also -- If you observe good calibration out to 1e-3 or so, it should be reasonable to assume that the false discovery rate is controlled (provided that there are some true signals among your discovery pairs).

[argschwind]

We observe severe miscalibration with the high moi setting compared to low MOI. We use the same covariates in both analyses, so not sure what's happening here.

Low MOI:

High MOI:

[timothy-barry]

I see. Yes, the high MOI will need to be looked into.