I have checked that all genes that are configured exist in our genome (HGNC)
The LOH check in immune escape is inconsistent (MACN < 0.3) with other LOH algo's in ORANGE (MACN < 0.5). The current GIE implementation is consistent with the GIE paper and vignette though, and according to Fran deliberately chosen to be conservative.
I have chosen for the immune escape to be omni-present (NotNull). I feel this is the fine choice, but does require some considerations:
If some genes that are checked in immune escape analysis but were not part of the panel that went into ORANGE, we will have missing data. Have decided to go with a WARN in this case, but may not be the cleanest.
If QC is e.g. FAIL_NO_TUMOR, it may be cleaner to have null immune escape instead of false everywhere. However, "false" is still not a wrong value in this case, so want to leave up to downstream (and in theory we could still find something in case of non-PASS QC and don't want to filter).
Note that in the report I blank out the immune escape table in case of QC failure!
Have added a screenshot of the new immunology page of COLO829 report for ease of review of the reporting changes.
See JIRA for design and rationale.
Some notes on implementation:
FAIL_NO_TUMOR
, it may be cleaner to have null immune escape instead of false everywhere. However, "false" is still not a wrong value in this case, so want to leave up to downstream (and in theory we could still find something in case of non-PASS QC and don't want to filter).Have added a screenshot of the new immunology page of COLO829 report for ease of review of the reporting changes.