Our integration test currently checks type assignments, but does not check that haplotypes are consistently called - because for a given patient, two different analysis methods may legitimately produce different haplotypes.

Haplotype regressions (parses that used to work, but don't after an update) are a common issue we run into. To mitigate this, we should at least ensure that the haplotype for each file is parsed consistently.

Steps:

1. For our current collection of integration test patients, for each PDF and XML file with Haplotype data, cache the Multimap<RaceGroup, Haplotype> from the ValidationModel parsed by the appropriate DonorFileParser. This is our baseline haplotype data.
2. Create a map of source files to cached baseline haplotype multimap.
3. In the parsing section of our IT, if a file has an entry in the haplotype map, run a method that checks if the parsed's haplotype multimap matches up with the cached version for that file.

Note: the haplotype check should only compare parsed to cached haplotypes if the haplotype frequeny tables are present, i.e. the HLAProperties.get().getProperty(NMDP_DRDQ_PROP) and HLAProperties.get().getProperty(NMDP_CB_PROP) properties are set

Desired menu entries:

• About - show app version, link to comp path, link to github
• Report an issue - link to google group

When reporting errors on patient loading, add link to e-mail the DonorCheck group, pre-populated with the stack trace of the exception.

See attached file: AGID359 HLA.pdf

Currently these windows don't respect screensize and thus go off screen.

Wrapping them in scroll panes should be sufficient to get them to display properly

Our lab is switching to a new SureTyper PDF format that we need to validate.

• After seeing that spaces are inconsistently distributed throughout typings (hooray) but whitespace is never a delimiter for alleles, before splitting into tokens, let's remove all whitespace. Then we can split on , and ; to look for alleles.

• Instead of skipping tokens that start with the locus, let's just remove the locus text from the token. Then if the token is empty we can skip it, otherwise parse it normally.

• I can't find the code line to reference because GitHub is not cooperating at the moment, but to accommodate the HLA-DR replacing HLA-DRB1 in the haplotype area, you can just add another mapping in the haplotype map, from HLA-DR to the same map mapped to by HLA-DRB1... I think? that way you don't need explicit checks for HLA-DR in the parsing logic

Currently hard-coded a removal of "+" which is not ideal.

Referenced here.

XML Sections to parse

It looks like the XML has an labAssignmentSection block at the top which includes the assigned typings. These can be read and passed to a ValidationModelBuilder to populate the core typing data.

For haplotype information, there is a testResultsSection that has all the possible alleles for each locus, which will be used to populate haplotype information.

Implementation Notes

• You can use the Score6 XML parser as a reference for how to read the desired XML sections.
• For reference, the XML parsing is done with JSoup. You're given the Document, so building the parser is all about finding the appropriate XML Elements, getting their text contents, and passing them appropriately to the builder.

Goals

• Assigned typing parsing works
• Haplotype parsing works

• The XML download steps have changed - we need to confirm them and update the tutorial. (From donor summary first tab?)
• HTML instructions need to specify: from donor summary, select crossmatch and then hla subtab
• Firefox/chrome clarifications need to be clearer
• Clarify that xml is only available after typing

Suretyper PDF missing C, DQA, DPB, DRB and DPA Loci when attempting to build.

• Identify reason they are missing
• Fix reason they are not being parsed
• Test fix on given samples
• Add sample to integration test
• Create new development installer

A general walk-through of the app would be nice

SureTyper PDF file will not import. Error message appears (see below)

Error message: Could not read donor data from PDF. Please notify the developers as this may indicate the data has changed.

New Score6 version is missing an indicator for undefined. This causes results that vary from old results as in the case of:
new set:

<alleleCombination1>
DQA1*01:03:01:01, 01:03:01:02-01:03:01:09, 01:10, 01:14, 01:15N, 01:17, 01:24
</alleleCombination1>
<serologicalCombination1>Null</serologicalCombination1>

vs old set:

<alleleCombination1>
DQA1*01:03:01:01, 01:03:01:02-01:03:01:09, 01:10, 01:14, 01:15N, 01:17, 01:24
</alleleCombination1>
<serologicalCombination1>-, Null</serologicalCombination1>

Need to:

• Find cause for dqa1 being missed
• Find solution to work for all cases: empty, only "Null", and partially "Null"
• Test solution for samples give to us
• Add to integration testing
• Build new development installer with issue fixed

• Split "assigned types" and "haplotypes" to defined components instead of lumping them all together in one class
• Document the Builder methods to better explain what should be passed as input