"Something went wrong during Hit Calling calculations."

[Klaas de Lint]

After uploading my data, starting the analysis and waiting for about 40 minutes I get this error message, is there any way to find out what exactly went wrong?

[Klaas de Lint]

Our local installation keeps crashing near the end of the analysis too, but not entering any guides as controls seems to have prevented the bug. Some methods will not work properly without the controls though, I'll try to include different sets and see if the analysis succeeds.

[Jan Winter]

Hi Klaas,

thanks for your email.

I realized that in some conditions the use of controls leads to a crash in the analysis.

I am yet not sure what causes this issue, but I suspect that it occurs once the controls have a too low read count or actually no read counts present for this.

As an example:

In one of my test data sets I deleted the controls from the read count data, but not from the Fasta -> Crash when entering the controls. CRISPRAnalyzeR uses the FASTA to determine if controls are present.

Did you directly start with FASTQ data or not?

I will have a closer look and would be happy if you can also send me some logs after you're done with the runs.

You can get them on the About CRISPRANalyzeR page on your local installation. In the web-demo, this is disabled.

Thanks for your help,

Jan

[Klaas de Lint]

Hi Jan,

To check if zeros in the read counts caused the problem, I uploaded read counts of a drug sensitizer screen (Untreated vs. Treated, 3 replicates each), and the online analysis crashed while calculating the BAGEL scores.

I added 1 to each count, but the analysis crashed at the same stage.

Next, I used the same count files with 1 added to each count, but I checked the box 'YES, remove sgRNAs with a read count LOWER or equal to' and set the value to zero, and this time the analysis proceeded, unfortunately the server disconnected shortly after while we were trying to download the report.

We'll try if our local install runs fine using these settings and check the logs if there are any issues. Best regards,

Klaas

[Jan Winter]

hi Klaas,

any news on this?

I guess it worked?

cheers

Jan

[Klaas de Lint]

Hi Jan,

The analyses run fine now, one issue we had running it on the dkfz server is that a server disconnect kept happening some time after the analysis has completed, while browsing the results. We got around this by exporting the HTML report upon completion of the anaysis.

At the moment we run Crispranalyzer on Google Cloud and this works great, so I don't know if the server disconnect is still an issue when running it at dkfz, I can try this again. Cheers,

Klaas

[Jan Winter]

Wow, good to hear!

Yeah our online demonstration might disconnect sometimes, as it sits behind a really strict proxy, that is not compatible to modern technologies such as data streaming.

Would you mind sharing a protocol or so how to host it on the google cloud or similar instances?

Some users like to do that and I do not have capacities to test this out.

Thanks a lot, I hope your results are satisfying :)

cheers

jan

[Klaas de Lint]

Sure, I'll ask my colleague Agata, who did the cloud installation, if she can describe how she managed.

We're very happy with the analysis results, Crispranalyzer is a great way to explore our screen data and compare the different analysis methods, keep up the good work! Cheers, Klaas

[Agata Szymanek]

Hi Jan!

I was in fact planning to write some small "how to" for our internal needs about setting CRISPRAnalyzeR on Google Cloud so of course I will be happy to share our way around it once it's ready :)

All the best!

Agata

[Jan Winter]

Amazing! Thanks in advance :)