Different outputs from Pooled vs individual amplicon analysis

[danjpri]

Hi,

Thanks very much for the software, and also in advance of any help you can provide!
I have been attempting to optimise the command for pooled analysis of a sequencing run containing 3 clones but have noticed that the data output is slightly different after the amplicon information is formatted in the .txt file, when compared to just listing the amplicons in the code.

Originally, I was using the following line:
Original amplicon line.txt
Then, in the pooled format:
Pooled line.txt

The .txt file with the amplicon information:
cl21_25_29_amplicon_crispresso2originalpeg.txt

Can you determine what is being missed in the pooled line that is causing a slightly different output?
Thanks very much again!

Best,
Daniel

[Colelyman]

Hi Daniel,

Thank you for using CRISPResso! The files look good to me, it may be helpful to figure out what is going on if you provide what is different in the output between the two commands. If you aren't comfortable sharing the full report, a good start would be describing which aspects are different between the reports (e.g. number of reads assigned to the amplicons, different quantification windows, ...).

Cheers,
Cole

[danjpri]

Hi Cole,

Thanks for the quick response. Here are the summary reports for the 2 different methods:

Seperate amplicon:
CRISPResso_quantification_of_editing_frequency.txt

Pooled:
SAMPLES_QUANTIFICATION_SUMMARY.txt

I initially noticed the difference in the nucleotide percentage tables, but can now see there are differences in # reads alligned and other parameters. Any idea what is differentiating the 2?

Thanks again!
Dan

[Colelyman]

Hi Dan,

Thanks for the output files and sorry for the delay. I believe the difference is subtle, and is related to how CRISPResso initially assigns the reads to the amplicons. When running CRISPResso with multiple amplicons, CRISPResso will align each read to each amplicon and assign the read to that amplicon that has the highest alignment score.

CRISPRessoPooled will first use bowtie2 to initially assign your reads to the amplicons, then refine the aligments for the reads assigned to each amplicon using the same alignment as above. Because bowtie2 has a less precise alignment (compared to the CRISPResso alignment which is an optimal alignment) there may be differences in how it assigns the reads initially. Especially because your amplicons have similar sequences bowtie2 may misclassify some reads.

If you are willing to share your input data we would be happy to take a further look (after the holidays) and see exactly what is happening. If you would rather not share it publicly, you can email me at cole@edilytics.com.

Cheers,
Cole