Changes between Initial Version and Version 3 of Ticket #1046


Ignore:
Timestamp:
May 7, 2018, 2:19:00 PM (7 years ago)
Author:
Nicklas Nordborg
Comment:

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #1046

    • Property Status new → assigned

  • Ticket #1046 – Description

    initial v3  
    11This is the second step in the genotype QC. Alignments that were flagged by the first wizard should be further examined and it should be possible schedule re-runs from RNA or re-extractions from the specimen.
    22
    3 The wizard will probably require different approaches for different problems:
     3It was not so easy to write a specification for this wizard. Instead, I will write something about the ongoing development as soon somethings seems reasonably stable to be useful.
    44
    5  * High HET: This is probably the easiest to handle. A re-run from RNA or re-extraction is scheduled. When the result arrive there are two outcomes:
    6   - HET has dropped: Disable the original alignment and accept the re-run.
    7   - HET is the same: Accept both.
     5'''A description of the development so far'''
    86
    9  * Qiacube mixup: Schedule re-extractions for samples that was part of the qiacube. We should expect pairs of HIGH/LOW mismatch. The problem can be solved by disabling alignments from the problematic qiacube and accepting the re-runs. Theoretically, the original alignments can be saved by a (complicated) renaming procedure. The secondary analysis part should be deleted and re-executed (since the original names may be part of file data).
     7The wizard will start out in a context-less mode which simply means that it will load all alignments from the '''Flagged alignments''' list and compare them to each other. In this mode it will make two simple checks:
    108
    11  * Libplate rotation: More or less the same actions as for a qiacube mixup, but it may be enough the re-run from RNA.
     9 1. Check the `QC_GENOTYPE_VERIFIED` annotation. This is very useful since it allows us to know alignments that we believe to be correct.
     10 2. Other alignments that has a HIGH or LOW mismatch against a verified alignment is marked as a `Verified mismatch`. This is an alignment that we believe to be incorrect.
     11
     12This will be listed together with information about the !QiaCube and !LibPlate information. The user may click on a !QiaCube or !LibPlate. This switches the wizard to a context-sensitive mode.
     13
     14In the context-sensitive mode the wizard will load all alignments (including non-flagged alaignemts) that are related to the context. For example, for a given !QiaCube run, the wizard will load all RNA items and follow them up to the Patient level and loading all alignments related to those patients.
     15
     16In addition to the two simple checks above, the wizard will now also make a third check that will try to locate a possible mixup in the current context. Here is an outline of the procedure:
     17
     18 * Look for a pair of alignments A1 and A2 where
     19   - A1 is in the current context
     20   - A2 is in another context or a VERIFIED alignment
     21   - there is a LOW mismatch warning between A1 and A2.  This indicates that A1 and A2 is from the same patient.
     22 * Get the patient from A2 and try to find another alignment A3 in the current context that has the same patient. A3 must not be a "Verified mismatch".
     23 * If we find it, this could mean that A1 is actually A3.
     24 * This may not be enough information to identify the problem. Typically, a VERIFIED alignment is needed to be sure.
     25