DPYD phenotype output discrepancy and genotype normalization behavior in PharmCAT 3.0.1

[ranijames]

Dear All,

Thank you for the wonderful tool.

We are currently using PharmCAT version:

pharmcat-3.0.1.jar

Our pipeline uses the PharmCAT outside-call TSV input (--po) because our upstream processing is based on hg19 reference data.

• Command line used
  /java17/bin/java -jar /pharmcat-3.0.1.jar -o /data/output -po /input/1221861.tsv -phenotyper -reporter

We observed unexpected behavior for the DPYD gene and would appreciate clarification regarding the genotype normalization and phenotype assignment logic.

Observation 1 — Phenotype shown as “See drug section”

For the following DPYD diplotype:

Reference/c.85T>C (*9A)

PharmCAT generates the following genotype summary output:
c.85T>C (*9A)/c.496A>G

with the phenotype column displayed as:

See drug section

(see Attachment 1)

However, according to the CPIC DPYD phenotype translation table:

DPYD phenotype translation table

allele with normal function/activity score 1.0 are expected to correspond to:

DPYD Normal Metabolizer

In our case, PharmCAT reports:

Two 1.0 (Normal function) alleles

but the phenotype field does not explicitly display:

DPYD Normal Metabolizer

Could you please clarify:

1. Why the phenotype column displays “See drug section” instead of “DPYD Normal Metabolizer”?
2. Whether this behavior is expected in PharmCAT 3.0.1 for DPYD HGVS-based outside-call inputs?

Observation 2 — Genotype normalization behavior

Our outside-call TSV input contains the genotype:

Reference/c.85T>C (*9A)

However, PharmCAT displays the genotype in the report as:

c.85T>C (*9A) (heterozygous)

(see Attachment 2)

It appears that PharmCAT internally normalizes or transforms the supplied genotype representation by:

removing the explicit Reference/ allele
converting the representation into a heterozygous HGVS-style notation

Could you please clarify:

1. Whether this normalization behavior is expected?
2. How PharmCAT internally interprets Reference/ alleles for DPYD HGVS inputs?
3. Whether the displayed heterozygous representation is functionally equivalent to the original outside-call TSV input (Reference)?

Thank you very much for your help and clarification.

[markwoon]

Could you please provide the full command line you are using? We are having trouble reproducing your results. If you are providing an outside calls file, please also upload it so we can reproduce.

[whaleyr]

I can reproduce the "Observation 2" result with just having DPYD Reference/c.85T>C (*9A) in an outside call file.

I cannot reproduce "Observation 1" results. Are you sure that one was done with just an outside calls file? If so, can you post the DPYD line of that file?

[ranijames]

The outfile used for first and secound attachments
test.tsv

DPYD    Reference/c.85T>C (*9A)

test1.tsv
DPYD c.85T>C (*9A)/c.496A>G

[whaleyr]

I can't reproduce your results. If I enter Reference/c.85T>C (*9A) in my outside call file I do NOT get c.85T>C (*9A)/c.496A>G like you say in Observation 1. Maybe your files got mixed up.

However, I can address your other questions.

The DPYD calling normalization is expected. See our DPYD docs page. PharmCAT normalizes Reference/c.85T>C (*9A) to c.85T>C (*9A) (heterozygous) because PharmCAT doesn't interrogate the entire DPYD sequence so we can't truly say the sequence is reference. The app just detects the changes at individual positions. Therefore, if we detect a single variation we can say it's heterozygous without making claims to the rest of the sequence. You can consider what is reported as equivalent to the value specified in the outside call file.

The "See drug section" value in the phenotypes column is a historical holdover that's no longer relevant. In short, there were specific scenarios where CPIC and DPWG phenotypes assignment may diverge for the same diplotype. You may not have one phenotype value that applied to all drugs+sources. You would have to look at the individual drug results in order to see what the phenotype value was for each drug+source. This will change because we no longer report on the DPWG function/phenotype assignment and should be able to report a single phenotype for each diplotype call.

[ranijames]

@whaleyr @markwoon Thanks for your detailed reply. Sorry for being late with responses.
I have one more question. According to your comment, there is no single genotype ↔︎ phenotype relationship; the phenotype depends on genotype, drug, and source. Is that correct to assume?

[whaleyr]

There is no table in the database that literally lists out a phenotype for every possible genotype. The gene_result_lookup table holds the association between the values needed to look up a phenotype and the phenotype value.

HOWEVER, there is a "view" called diplotype that does the proper genotype->function->phenotype or genotype->activity score->phenotype process for you.

For example, you can see this in the API by doing something like this: https://api.cpicpgx.org/v1/diplotype?diplotype=eq.c.85T%3EC%20(*9A)%2FReference