Clinical-Genomics / cg

Glue between Clinical Genomics apps
7 stars 2 forks source link

Alert the user of unstarted analysis due to `has too few reads` #3392

Open RasmusBurge-CG opened 2 months ago

RasmusBurge-CG commented 2 months ago

As a user of TB, I want alerted to the fact that a sample did not start due to not meeting the required amount of reads, So that we don't delay delivery on reruns.

The full story, we were contacted by the customer asking where the analysis where (#582062,#256210).

We saw the following: ACC6912A16 : Reads: 601768782 < minimum reads for app tag 0.92 660000000 = 607200000 ACC12275A1 : Reads: 601653018 < minimum reads for app tag 0.92 660000000 = 607200000

Looking in the Crown mail we could see that these samples were never started, too few reads. The error is pretty much silent, it would be perfect if we could be notified in some way, TB is mostly used to keep an eye on stuff, could be nice to have it there.

2024-07-03 13:17:24 [hasta.scilifelab.se](http://hasta.scilifelab.se/) cg.services.sequencing_qc_service.quality_checks.utils[41753] WARNING Sample ACC6912A16 has too few reads.

Acceptance Criteria

Notes

Implementation plan

islean commented 2 months ago

So apart from the mails and logs you can also see it in the order view for the order the sample was included in:

image

We can have a discussion about what alert would be the most valuable.

RasmusBurge-CG commented 1 month ago

After a discussion with @islean and @seallard. I want to be able:

RasmusBurge-CG commented 1 month ago

Hi again,

We have been thinking (together with @Karl-Svard). These samples are flagged as FAILED SEQ. QC, although they most likely did not fail under the criteria used previously. This discrepancy is probably due to a change in the application tag criteria. Therefore, one could argue that CG is not backwards compatible.

To remedy this, we should expand on an existing feature, Application Version. Why not include all the criteria there, such as the minimum number of reads and other relevant metrics? Samples are already assigned an Application Version. We should then use the originally assigned Application Version when restarting a case. An additional benefit could be that we automatically disqualify samples with an outdated application tag for pipelines that might require specific criteria.

islean commented 1 month ago

Hi again,

We have been thinking (together with @Karl-Svard). These samples are flagged as FAILED SEQ. QC, although they most likely did not fail under the criteria used previously. This discrepancy is probably due to a change in the application tag criteria. Therefore, one could argue that CG is not backwards compatible.

To remedy this, we should expand on an existing feature, Application Version. Why not include all the criteria there, such as the minimum number of reads and other relevant metrics? Samples are already assigned an Application Version. We should then use the originally assigned Application Version when restarting a case. An additional benefit could be that we automatically disqualify samples with an outdated application tag for pipelines that might require specific criteria.

We are planning on storing the aggregated sequencing qc status for cases in status_db. If these are old cases as well, this would solve your highlighted issue since the case would already be marked with a pass. If this is a new case with old samples, I guess we would need to refine how to solve it though.

beatrizsavinhas commented 1 month ago

After a discussion with @islean and @seallard. I want to be able:

  • to filter statuses in the column Analyses in the Orders view in TB.
  • to see if an analysis failed to start and why in the Analyses view. (comment: I know that this might be difficult due to the fact that the analysis did not start, and thus the analysis does not exists. However, wishes do not always need to be logical)

Regarding these wishes, I just wanted to point out that these objectives are included in the project proposal Improve filtering functionality in Trailblazer. Will specify as well that it should be possible to visualise the reason for the status FAIL SEQ. QC! Feel free to add or suggest features to include in this project.


I also agree with the idea of expanding the use of Application Version to make CG more backwards compatible. Me and @eliottBo have precisely found another case with the same problem. This solution won't solve problems with existing samples, but could prevent similar problems in the future.