Coverage in FACS based screens

[Ashwin Iyer]

Hi all,

Can someone please explain why maintaining coverage is important in cells post-sorting. I have a sample where I want to sort for top and bottom 25%. Pre-sort the coverage is maintained at 450x and the cells were infected at an MOI of 0.3 and grown for 7 days before taking it for FACS sorting.

Should I be worried about the number of cells I get out of the FACS instrument and any tips on how to ensure adequate coverage.

Thanks and Best Regards,

Ashwin Iyer

[Hamish McWilliam]

Hi Ashwin

You want to keep the unsorted library at the coverage that you transduced the cells at - 450x. But when you sort, you will get a smaller sample of this library, and this would be the new number that you should keep these sorted pools at. E.g. if you get 5 million cells from top and bottom sorts, then keep these pools at 5 million cells or higher. 

Any time you go below the original coverage level - whether the original transduction or the sorted pools - then you will lose coverage.

Cheers,

Hamish

--
You received this message because you are subscribed to the Google Groups "Genome Engineering using CRISPR/Cas Systems" group.
To unsubscribe from this group and stop receiving emails from it, send an email to crispr+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/crispr/7945c554-1741-44ae-a7a3-09f151f82c0fn%40googlegroups.com.

[Ashwin Iyer]

Hi Hamish,

Thanks for your response. So assuming I need to obtain 10 million cells post sort in each population (Hi and Lo) to obtain a coverage of 450X, so should I back calculate and take that many cells for sorting? The reason I ask is if you start with 450x coverage and your sorted samples do not have the same coverage, won't the analysis be noisy? Please let me know your thoughts.

Best,

Ashwin Iyer

[Hamish McWilliam]

Hi Ashwin,

Perhaps it's best to give an example. Here's some made up numbers:

1. Your library is 100k sgRNAs. You transduce 150 million cells at 0.3 MOI. Then you kill untransduced so you're left with 45 M cells = 450x coverage

2. You sort 50M cells. You sort the top/bottom 25%, so you get back 8M cells (slightly less than expected due to inefficiencies). Your sorted populations don't require the full 450x coverage, as you've selected them now. The sorted pops are now 80x coverage, but this doesn't really apply here - you actually don't want them to cover the full library; you want them to be enriched for certain guides.

3. Maintain the original unsorted library at 450x and the sorted at their 80x (8M) size. Or greater. But if you go lower, you will lose coverage. Sort again or just isolate the DNA as soon as you're ready.

Happy to hear other's suggestions - this is how we do it and how I interpret the original papers.

Cheers

H

To view this discussion on the web visit https://groups.google.com/d/msgid/crispr/327ee023-42dc-44fb-88cd-291fbe94a641n%40googlegroups.com.

[Ashwin Iyer]

Thank you Hamish for the detailed explanation. So in your experience doing these FACS screens, I have two follow up questions-

1) whats the minimum coverage you would shoot for if you want a good number of hits. My understanding is as the coverage goes down, the hits go down too.

2) Also, you would compare hi vs lo for analysis right? I would assume thats the best comparison due to comparable coverage. It could be context specific but in a generic FACS based screens like the one you explained, how would the analysis be if I compared Hi and pre-sort vs Hi and Lo? 

Thanks for your time.

Best,

Ashwin Iyer

[Ashwin Iyer]

Just to clarify my first question-

1) Minimum coverage in the sorted population. The starting coverage would be 450x. 

Thanks,

Ashwin Iyer

[Hamish McWilliam]

1) whats the minimum coverage you would shoot for if you want a good number of hits. My understanding is as the coverage goes down, the hits go down too.

As I said, coverage is not imporatnt in the sorted populations.

For the sorted populations, how much of the population you sort this is really variable. We typically sort the top/bottom 5% but I'm not sure what is the best; it is depedent on the question and biology of the screen.  I go for 5% sort because it is more likely to enrich for real hits imo. I lean towards wanting only a few 'real' hits over lots of false positives. Typically validating hits takes a lot of time so I would aim for fewer real hits.

 

2) Also, you would compare hi vs lo for analysis right? I would assume thats the best comparison due to comparable coverage. It could be context specific but in a generic FACS based screens like the one you explained, how would the analysis be if I compared Hi and pre-sort vs Hi and Lo? 

We compare the unsorted to the high, or unsorted to the low. Imo this is the best comparison. Perhaps also try comparing low to hi but I found more false positives this way. It makes more sense to me if you ask; what sgRNA are enriched in a sorted population relative to the whole library? Sorting can introduce a bottle neck so there may be lots of false positives; it is better to rule these out prior to validating them individually.