GeneyxPgxParser
Converts pharmacogenomics caller output — PharmCAT, pbStarPhase, PyPGx, the ONT PGx workflow, a Thermo Fisher PharmacoScan report, or a plain Gene/Diplotype table — into the JSON format Geneyx Analysis consumes for PGx reporting.
Install it next to your own PacBio, ONT or array pipeline, run it on the output, and upload the result to Geneyx.
pip install GeneyxPgxParser
geneyx-pgx-parser HG002.pharmcat.report.json HG002.pbstarphase.json --output-json HG002.geneyx-pgx-results.json
Output format
A JSON array, one entry per gene - every gene the input files call, nothing dropped:
[
{
"Id": "CYP2C19",
"Name": "CYP2C19",
"StarAllele": "*1/*1",
"Activity": "Normal Metabolizer",
"HaplotypeAlleles": null
},
{
"Id": "IFNL3",
"Name": "IFNL3",
"StarAllele": "Reference/rs12979860",
"Activity": "Unfavorable Response",
"HaplotypeAlleles": [
["Reference"],
["rs12979860", "rs8099917", "rs12980275"]
]
},
{
"Id": "HLA-DRB1",
"Name": "HLA-DRB1",
"StarAllele": "*04:01/*07:01",
"Activity": null,
"HaplotypeAlleles": null
}
]
Id/Name— the gene symbol. Both are required; the Geneyx report keys its result set onIdand filters onName.StarAllele— the diplotype (*1/*2), its two alleles in PyPGx's order whichever order the caller wrote them in (see Merging rules).Activity— the phenotype / metaboliser status (Normal Metabolizer), ornullwhen it is not known.HaplotypeAlleles— every allele the caller found on each of the two haplotypes, in the caller's order, ornull. Only PyPGx lists them (Haplotype1/Haplotype2); itsGenotypenames each haplotype by one allele only, the first in its allele table, so in the IFNL3 example above rs8099917 and rs12980275 were called but appear nowhere else.StarAlleleis unchanged by it: the report looks results up by that exact string. Two things to know when showing it. The lists are in PyPGx's phase order (Haplotype1,Haplotype2), which need not beStarAllele's order: CYP2C19*1/*17can come with[["*17", "*1"], ["*1"]]. And a list holds every allele whose definition the haplotype matches, so an allele that another contains appears beside it (*1with*17). A gene PyPGx calls from copy number alone (GSTT1, UGT2B17) has empty haplotype cells andnull.
These names are matched case-sensitively on the Geneyx side. Do not rename them.
HaplotypeAlleles and UnsupportedAlleles are additions the Geneyx side
does not read yet; its JSON reader skips members it does not know, so they are
safe to carry. Whether the Geneyx PGx report will show a result is not part of
the result array: InReport and ReportNote are in the provenance file, per
gene (see below).
A second file, <output>.provenance.json, is written alongside the result. It
records the caller versions, the mapping-table versions used, per-gene warnings,
and any term that had no Geneyx mapping. It is not part of the upload contract —
it exists so a report can be traced back to what produced it. Each gene carries inReport and reportNote: whether the Geneyx PGx
report will show the result and, if not, why (the gene is not in the report's
dataset, has no drug there, or its StarAllele is not one of the dataset's
diplotypes; a no-call or an unsupported allele says so). A gene's warnings
also say when the Geneyx PGx report cannot show its StarAllele (the report
looks a result up by that exact string), or shows a different phenotype for it.
Each gene also carries vendorDiplotype: the winning caller's diplotype exactly
as it wrote it (*2.001/*4.015 from pbStarPhase, a Thermo token, an HLA allele
with all its fields), before any rule turned it into StarAllele. The suballele
is the only place two *4 chromosomes can later be told apart, and PharmVar
assigns function at the core level, so it is kept here and never matched on.
Each gene also carries coverageCaveats: what the caller could not see
for it, which the reported call does not rule out - PharmCAT's
Missing_Positions / Uncallable_Haplotypes (the matcher TSV) and
uncalledHaplotypes / variants without a call (the JSON), Thermo's
Unknown Call alternatives (GAN-9404). For NA19207 that is CYP2C9 *6/*25/*85,
G6PD (9 variants), NUDT15 *6/*9/*18, DPYD *3/*7 and RYR1. The result is
unchanged by them. The sidecar's top level has samples, the sample each
input file names (PharmCAT's title, the matcher TSV's Sample column, the
ONT CYP2D6 report's Sample row, PyPGx's first column, Thermo's CEL file),
and warnings for the run: a sample-ID mismatch between the files, or
between a file and the sample its name is for (NA19207's matcher TSV says
7NA19207). Nothing is dropped for it.
A PyPGx gene also carries phaseExtended: true when a called allele has
every defining variant absent from the 1KGP panel PyPGx phases with, so Beagle
dropped them and PyPGx placed the allele by its phase-extension heuristic
(pypgx_panel_gaps.json, a measured list per PyPGx release); false for any
other PyPGx call; null for a caller whose phase is not PyPGx's. The result
is unchanged by it.
Supported inputs
| Input | Detected by | Provides |
|---|---|---|
*.pharmcat.report.json |
name or pharmcatVersion |
diplotypes, phenotypes, tool + data versions (PharmCAT 2.x and 3.x) |
*.pharmcat.phenotype.json |
name or geneReports |
diplotypes, phenotypes |
*.pbstarphase.json |
name or gene_details |
diplotypes; alone (PacBio WDL v4), the primary caller with PyPGx phenotypes |
*.pbstarphase.tsv |
.tsv header |
as the JSON (PacBio v4 summary) |
<GENE>.tsv (PyPGx) |
.tsv header |
diplotype + phenotype, one gene per file |
*.pharmcat.tsv (ONT) |
.tsv header |
diplotypes only (PharmCAT matcher; no CYP2D6) |
CYP2D6/report.tsv (ONT) |
Sample first row |
CYP2D6 diplotype only |
*_phenotype.rpt (Thermo) |
.rpt or Known Call |
diplotype + phenotype (PharmacoScan plate report; --sample) |
*.csv / *.json table |
fallback | Gene + Diplotype (+ optional Phenotype) |
The minimal table is the upload format: a header with Gene and Diplotype
is enough. Column names are matched loosely (gene, geneSymbol, symbol;
diplotype, genotype, starAllele).
Why two files are better than one
PharmCAT only reports genes that have a CPIC or DPWG guideline. In a PacBio
HG002 run, pbStarPhase called 35 genes and PharmCAT reported 22. Passing the
pbStarPhase summary alongside the PharmCAT output adds the other 13 — NAT2,
CYP1A2, CYP2A13, CYP2C8 and the HLA genes beyond HLA-A/B — with a diplotype and
no phenotype; the provenance file's inReport says which of them the report
can show.
Where both files describe the same gene, PharmCAT wins: it normalises
PharmVar sub-alleles (*4.015) to the core alleles (*4) that Geneyx uses.
pbStarPhase on its own (PacBio WDL v4)
The HiFi human WGS WDL v4 dropped PharmCAT, so a v4 run has only the
pbStarPhase output. Then pbStarPhase is the primary caller: every gene it
calls is reported, sub-alleles stripped and names put in the Geneyx (PyPGx)
vocabulary, and Activity is PyPGx's phenotype for the diplotype — what a
PyPGx run would report. On HG002 that is all 35 genes, 17 with an Activity.
Up to 0.1.5 the parser treated pbStarPhase as PharmCAT's supplement even when
it came alone, and a v4 sample came out with 4 genes and no Activity. With
PharmCAT in the run, PharmCAT's call wins for the genes it reports, and
pbStarPhase's phenotype is never attached to PharmCAT's diplotype.
ONT PGx output
The ONT PGx workflow writes its calls to two files, and both are needed:
<sample>.pharmcat.tsv holds every gene except CYP2D6, and
CYP2D6/report.tsv holds CYP2D6. Pass both, and they merge into one array:
geneyx-pgx-parser NA19207.pharmcat.tsv NA19207/CYP2D6/report.tsv --output-json NA19207.geneyx-pgx-results.json
The workflow runs only PharmCAT's allele matcher, not its phenotyper, so
neither file carries a phenotype. The Activity is looked up in CPIC's own
diplotype table - the table PharmCAT's phenotyper reads - packaged as
Mappings/cpic_diplotypes.json. The lookup tries both allele orders, writes
three or more copies the way CPIC does (*2x3 -> *2x≥3), and falls back to
the haplotype columns where the matcher TSV abbreviates (CFTR). Which CPIC key
answered, and its activity score, go in the provenance sidecar.
DPYD is usually unphased: PharmCAT lists the variants it found rather than a
diplotype, and they are reported as-is, ;-separated
(c.85T>C (*9A);c.557A>G). Two variants can only be one on each strand, so
the pair is looked up as a diplotype; three or more need CPIC's lowest-two
rule and are left without an Activity rather than guessed.
Genes the table cannot answer keep Activity: null: CYP3A4 and CYP4F2 have
no CPIC diplotype table, G6PD needs the sample's sex, and IFNL3 / VKORC1 are
genotype-guided as on every other path.
--input-dir does not descend into CYP2D6/, so name the report explicitly.
Keeping the CPIC table current
CPIC revises its tables in place - SLCO1B1 *41 went from uncertain to no
function on 2025-09-23, so NA19207's *37/*41 is Decreased Function here but
Indeterminate from PharmCAT 2.15.5, whose data predates the change.
The table is kept current in pgx-metadata, where
Test_CpicLiveTable.py fetches CPIC and fails while the table disagrees
with it, listing every changed, added and removed diplotype and every
phenotype term with no Geneyx term yet. The refreshed table reaches this
package as a new mapping bundle.
Running PharmCAT's phenotyper instead
The same Activity can come from running the rest of PharmCAT on the
workflow's <sample>.match.json, with the CYP2D6 call handed back to it as an
outside call. Use the same PharmCAT version as the workflow (wf-pgx v0.1.7
ships 2.15.5):
geneyx-pgx-parser NA19207/CYP2D6/report.tsv --pharmcat-outside-call NA19207.outside.tsv
pharmcat -phenotyper -pi NA19207.match.json -po NA19207.outside.tsv -reporter -reporterJson -o out -bf NA19207
geneyx-pgx-parser out/NA19207.report.json --output-json NA19207.geneyx-pgx-results.json
The report.json is the same file the PacBio path produces, and PharmCAT's
phenotypes are used as-is - no CPIC lookup. Unphased DPYD keeps its
;-separated variants as StarAllele and takes the phenotype PharmCAT gives
the pair.
Thermo Fisher PharmacoScan
Axiom Analysis Suite's apt-pgx-translation writes one *_phenotype.rpt per
plate, so the file holds every sample on it while the Geneyx result is per
sample. Name the sample, by CEL file name or its stem:
geneyx-pgx-parser 2026-02-18-PharmacoScan24F_phenotype.rpt --sample HG03225 --output-json HG03225.geneyx-pgx-results.json
A file with one sample needs no --sample; a file with several and no choice
is refused with the list of samples it holds (GAN-8867).
What the parser does with the report:
Known Callis the diplotype; Activity is PyPGx's phenotype for it. Thermo'sPhenotype Callis only a cross-check; see below.- Allele names are translated token by token through
thermo_alleles.json, built from the SNP-level Thermo ↔ PyPGx dictionary on GAN-8867 and the diplotypes the Geneyx model was seen to accept.WT,c.=andrs2231142GbecomeReference;*28+60+80+93becomes*80+*28; a male's G6PDNullbecomesMALE;MTRNR1 Null/m.=becomesMT-RNR1 Reference/Reference. The table names Geneyx core alleles, so a PyPGx compound such as*10+*4is written*4. - A token the table does not know is kept as written and listed in the
sidecar's
unmappedTerms. It is a genomic observation the Geneyx model does not name yet, not a vocabulary term that could pass for one: RYR1c.6178G>T, a CPIC Level A malignant-hyperthermia variant, is reported asc.6178G>T/Referencerather than vanishing intoReference/Reference. - Ambiguous phase is reported, not resolved. A
,-separatedKnown Call(*1/*14,*4/*37) is Thermo saying it could not phase the sample; the candidates are emitted;-separated, like unphased DPYD on the ONT path, and the gene is flagged in the sidecar. Candidates that name the same Geneyx diplotype (*1A/*7A,*1B/*7A+1B→*1/*7) collapse to it. - Copy-number gains are kept as
(*1/*10)xNwith a warning. Thermo does not say which haplotype is duplicated, so nothing is guessed. - Rows with no
Known Call(UNDH,CN_HybridLoss) are written as no-calls (StarAllele: null), with the interpretation code and theUnknown Callcandidates in the provenance. Up to 0.1.7 they were left out. Unknown Callbeside aKnown Callis a coverage caveat. Control1's CYP2D6*1/*29stays*1/*29NM; Thermo's alternatives with a haplotype the chip cannot resolve (*2/UNK,*34/UNK,...) go to the sidecar'scoverageCaveatsfor the gene, and weigh nothing (GAN-9404).- A copy-number code the
Known Callleaves out (CN_HybridGainon*2/*4) keeps the call and adds a warning that it may not describe every gene copy. - Every gene the chip calls is reported (ALDH2, FMO3, TNF, UGT1A9 …), with
inReportfalse(provenance) where the Geneyx PGx report has nothing for it. OnlyCYP2CRS12777823is left out: it is a carrier for the rs12777823 warfarin variant rather than a gene (ignored_genesinthermo_alleles.json). On the 24F panel that is 31 of the 32 reported genes.
Where Activity comes from: PyPGx's phenotype step
The Geneyx report is built on PyPGx. A Thermo report has a diplotype but not
PyPGx's phenotype for it, so that step is run here: once Known Call is
translated into Geneyx (PyPGx) allele names, Activity is what PyPGx's
predict_phenotype gives that diplotype — read from
Mappings/pypgx_phenotypes.json, which pgx-metadata generates from the PyPGx
release the pipeline runs (0.27.0).
A Thermo sample therefore reads as a PyPGx sample with the same diplotype
would:
- Diplotype genes (CYP2C19, TPMT, SLCO1B1, G6PD, CFTR, RYR1, …) take
PyPGx's diplotype table; a diplotype it does not list is
Indeterminate, as in PyPGx. - Score genes (CYP2D6, CYP2C9, DPYD) add the two alleles' activity scores
— copies multiply,
+tandems add — and take PyPGx's ranges; an allele of uncertain function isIndeterminate. - Genes PyPGx gives no phenotype (NAT1/NAT2, COMT, CYP1A2, GSTM1, VKORC1, …) get no Activity, as a PyPGx run leaves them.
- No guessing. An unresolved phase (
*1/*14;*4/*37) or an unplaced copy-number gain ((*1/*7)xN, read with every copy count that could change the answer, not only two or three) is reported only when every possible reading gives PyPGx the same phenotype; a diplotype with an allele the Geneyx model cannot name gets none. The provenance sidecar says which and why.
Thermo's own Phenotype Call (NM, NF, MHS, …) is converted per gene
through phenotype_calls in thermo_alleles.json and only compared: where
it differs from PyPGx, the gene carries a warning naming both. They differ
often — Thermo's research-use metabolizer table is coarser than PyPGx's (CFTR
NF vs PyPGx Indeterminate, G6PD Deficient vs Deficient with CNSHA,
RYR1 Indeterminate vs Malignant Hyperthermia Susceptibility for
c.14387A>G) — which is why it is not the reported value. The reconciliation
report lists every such difference on its Thermo vs PyPGx sheet.
Worth knowing when reading the output: for the twelve panel genes whose Geneyx
model carries only an Indeterminate activity (COMT, CYP1A2, CYP2A6, CYP2C8,
CYP2E1, CYP3A4, CYP3A7, CYP4F2, GSTP1, NAT1, NAT2, UGT2B7), the report will
return Indeterminate however good the diplotype is: PyPGx has no phenotype
for them, and Thermo's own call is not used. That is a gap in the Geneyx model's activity tables,
not in this parser, and it is the same gap GAN-8867 records for GSTM1.
Checking the Thermo tables against PyPGx
Both halves of thermo_alleles.json (allele tokens by variant content, and
every metabolizer diplotype's Activity) are checked against the PyPGx release
the Geneyx pipeline runs by tools/thermo_reconcile.py in pgx-metadata,
together with Thermo's library files and PyPGx's tables it reads, kept
byte-exact under sources/ there. A new Thermo library or PyPGx version is
brought in, reviewed and approved there; this package receives the result as
a mapping bundle. A report records its own library as thermoLibrary in the
provenance sidecar (read from its #%MetabolizerFile= header), and a report
made with a different library than thermo_alleles.json names is still
translated, with a warning on every gene.
Merging rules
- One entry per gene. PharmCAT emits a gene report per phenotype source, so the same 22 genes appear under both CPIC and DPWG. One source is preferred and the other is consulted only as a fallback — see below.
- One vocabulary for every caller. The report shows
StarAlleleandActivityverbatim and the PyPGx pipeline writes PyPGx's own names, so every caller's output is held to PyPGx's allele names and phenotypes: an Activity outside its gene's PyPGx phenotypes is not reported (PharmCAT's DPWGNormal Metabolizerfor CYP3A4), PharmCAT's ABCG2rs2231142 reference (G)is writtenReference, and DPYD'sc.1627A>G (*5) (heterozygous)label is writtenReference/c.1627A>G (*5), as PyPGx writes them. - One spelling for every diplotype. The two alleles are written in PyPGx's
allele order, as a PyPGx run writes them:
*4/*2becomes*2/*4,c.1627A>G (*5)/ReferencebecomesReference/c.1627A>G (*5), and each candidate of an unresolved call is ordered on its own. An allele PyPGx does not define keeps the caller's order. A male's single G6PD allele is written withMALE(MALE/A- 202A_376G), and a sub-allele number is dropped before a copy number too (*3.001xN→*3xN). The phenotype is unaffected: it is looked up in either order. - A no-call is written, never dropped. A gene the caller names but did not
call — an empty cell, pbStarPhase
./., PharmCATUnknown/UnknownorNo Result, ONTIndeterminate, a Thermo row with noKnown Call(UNDH,CN_HybridLoss) — is written withStarAllele: null,inReport: falseand areportNotestartingno call:in the provenance; the caller's own word is kept asvendorDiplotypeand the reason is in the provenance. A partial call (*1/.) is kept as written, with no Activity. Where the preferred caller made no call and another caller in the run did, the other call is reported, with a warning saying so. PyPGx's ownIndeterminateis passed through unchanged, as every PyPGx value is. - Nothing is called that the caller did not call. Several candidate
diplotypes (pbStarPhase's list, Thermo's
,, PharmCAT's source diplotypes) are all reported;-separated, never the first one alone. - Several possible diplotypes are resolved or Undetermined, never dropped
(GAN-9404). When every possible diplotype gives the same phenotype it is
reported: Thermo's
*1/*7,*5/*6(CYP2B6) is Intermediate Metabolizer either way, with the notation kept; two unphased DPYD variants (c.85T>C (*9A);c.557A>G) can only be one on each strand and are written as the pair the report lists,c.85T>C (*9A)/c.557A>G. Where PyPGx has no phenotype for the gene (CYP4F2, NAT2) the report dataset decides, each candidate looked up there. When the possible diplotypes give different phenotypes, or one of them has none -*1/*14;*4/*37(SLCO1B1: Normal Function or Indeterminate),(*1/*10)xN(no activity score without the copy number), NAT2 candidates only one of which the dataset knows -ActivityisUndetermined: the gene is reported with the caller's notation, nothing is chosen, and no phenotype-dependent recommendation applies. A single call with an allele Geneyx cannot name is a different case (GAN-9415) and is not Undetermined. The report looks a result up by its exactStarAllele, soinReportisfalsefor;notation and thereportNotesays so with the Activity. - Copy number, hybrids and losses. Deletions (
*5), gains (*2x2,*1x3) and hybrid tandems (*36+*10,*68+*4) are scored as PyPGx scores them; pbStarPhase's sub-alleles are dropped from each part of a tandem (*36.001+*10.001→*36+*10). A copy number left open —*2xN,*2x≥3, Thermo's unplaced(*1/*10)xN— is read at every count it could be, and gets an Activity only when they all agree (*1/*2xNis Ultrarapid whatever N is;*10x≥3/*10is Intermediate, Normal or Ultrarapid, so none). A Thermo copy-number code itsKnown Calldoes not spell out (*2/*4withCN_HybridGain, a partial duplication) isUndeterminedwhen any allele is not the reference - the variant may sit on the partial copy - and the call stands when they all are.CN_HybridLoss,UNDHandCN_Error(genotypes from more than one copy-number state, even beside a Known Call) are no-calls whose reason is in the provenance;CN_NoCall(no copy number reported) keeps the call and records a coverage caveat. An unplaced or open copy number isUndetermined, neverIndeterminate: that word is the report dataset's own phenotype term, not this parser's. - Unmapped terms are dropped, not passed through. A PharmCAT phenotype with
no Geneyx equivalent yields
Activity: nullplus a warning in the provenance file, rather than an unrecognised term that looks valid to the report. - Genotype-guided genes (IFNL3, VKORC1) take their
StarAllelefrom the rewritten allele names rather than from the phenotype. - HLA alleles are truncated to two fields (
*01:01:01:01→*01:01). HLA-A and HLA-B are reported by risk-allele presence (hla_risk_allelesingene_rules.json:*31:01;*15:02,*57:01,*58:01) on the caller's diplotype, whichever caller:Increased Risk - *15:02, *58:01 positive, orNormal Risk - *15:02, *57:01, *58:01 negative. The caller's own phenotype is not read, and the pair is kept as called. An allele too short to tell from a risk allele (*57,*57:XX) makes the resultUndetermined, as does a risk allele with an expression suffix (*57:01:01:02N, null: pending clinical review) and several pbStarPhase diplotypes (a/b;c/d) that do not all give one result. The report still resolves by exactStarAllele, so a pair its dataset does not list isinReport: falseuntil the dataset or engine changes (GAN-9412); the Activity is right in the output meanwhile. - Single-allele genes (MT-RNR1) are one call, written as PyPGx writes it:
Reference/Reference,m.1555A>G/m.1555A>G(PharmCAT's1555A>Gis renamed).
Choosing a phenotype source
PharmCAT phenotypes come from two guideline bodies, CPIC and DPWG, and they do not always agree. Pick one:
geneyx-pgx-parser HG002.pharmcat.report.json --phenotype-source DPWG
The default is CPIC (the first entry of phenotype_source_preference in
gene_rules.json). The choice only reorders the sources — it never narrows the
output. When the chosen source has no usable phenotype for a gene, the next
source in the default order is used instead, and the substitution is recorded:
{
"Id": "CYP2B6",
"phenotypeSource": "CPIC",
"warnings": ["no usable phenotype from DPWG; taken from CPIC"]
}
"No usable phenotype" covers three cases: the source reported none at all
(HG002 CPIC: CYP3A4, CYP4F2, VKORC1), it reported a no-call token such as
n/a (HG002 DPWG: CYP2B6), or its term has no Geneyx mapping yet (HG002 DPWG:
DPYD 2.0 (Normal Metabolizer)).
The source selection affects Activity only — StarAllele is identical either
way, since both sources describe the same call. The preference actually used is
written to the provenance file as phenotypeSourcePreference.
PyPGx, and callers whose terms are already Geneyx terms
The Geneyx PyPgx pipeline has produced this JSON format since it shipped, by
reading PyPGx's per-gene data.tsv and writing its Genotype and Phenotype
columns straight through. Those values are the Geneyx vocabulary, so mapping
them would change clinical output rather than normalise it — phenotype_terms.json
maps none of PyPGx's G6PD, MT-RNR1 or CFTR terms, and genotype_terms.json
knows only PharmCAT's allele names for the genotype-guided genes.
PyPGx is therefore listed in gene_rules.json under passthrough_sources: its
calls skip term mapping, sub-allele stripping and HLA truncation entirely, and
the sidecar records which callers were passed through:
{ "passthroughSources": ["pypgx"] }
Removing pypgx from that list is a clinical behaviour change gated on the
GAN-8770 mapping sheet, not a refactor.
Each PyPGx file covers one gene, and the gene comes from the file name
(CYP2C19.tsv), because PyPGx's own table has no gene column. A table still
named data.tsv needs a Gene column, or parsing fails rather than reporting
a gene called "data". Point the parser at the whole results folder:
geneyx-pgx-parser --input-dir /path/to/pgx/results --output-json sample.geneyx-pgx-results.json --gzip
Mapping tables
The clinical vocabulary lives in GeneyxPgxParser/Mappings/ as versioned JSON.
The tables are owned by pgx-metadata (Azure DevOps), where reference
sources are snapshotted, mapping proposals are reviewed and approved, and a
signed, versioned mapping bundle is built. This repository pins one bundle
in bundle.lock; the packaged tables are exactly that bundle's, and
Test_MappingBundle.py fails if a table is edited here. To adopt a new bundle:
python scripts/sync_bundle.py --from <pgx-metadata bundle folder> --public-key bundle.pub
and open one pull request with bundle.lock and the tables it changed.
pgx-metadata runs the parser against its candidate tables by naming their
folder in GENEYX_PGX_MAPPINGS_DIR; unset, the packaged tables are used.
| File | Contains |
|---|---|
gene_rules.json |
structural rules — source preference, passthrough callers, HLA/sub-allele handling |
phenotype_terms.json |
PharmCAT phenotype term → Geneyx Activity term |
genotype_terms.json |
allele/diplotype rewrites for genotype-guided genes |
thermo_alleles.json |
Thermo PharmacoScan allele token → Geneyx allele name, and phenotype code → CPIC term, per gene |
phenotype_terms.json is built from the approved crosswalk
(PyPGx_PharmCAT_phenotype_crosswalk 15.07.2026.xlsx, sheet
Phenotype Crosswalk (SD), linked from GAN-8770): its Rule class column
decides where a row lands, EXACT in the passthrough block and TRANSFORM in
the per-gene block.
thermo_alleles.json is generated, not hand-written: its source is the
Dictionary sheet of PGx_Thermo_PyPGx_Reconciliation_6.xlsx on GAN-8867
(SNP-content matching of every Thermo allele against PyPGx's definitions),
checked against the diplotypes the Geneyx model accepted in
PharmacoScan_diplotype_ingestion_report.xlsx. The handful of hand overrides
(CYP2C19 *1, the UGT1A1 TA-repeat tokens, G6PD Null, MT-RNR1) each carry
their reason in the file. It is marked DRAFT until it has clinical sign-off.
A
nulltarget means the crosswalk has no approved term, not that the row is outstanding work. HLA-A and HLA-B carry nulls the parser no longer reads: their Activity is the risk-allele rule's (see HLA above). The genotype-guided genes ingenotype_terms.jsonare still unfilled, because their table is a different sheet. Run the parser and readunmappedTermsin the provenance file to see what your data actually hit.
Development
pip install -e .
pytest GeneyxPgxParser/Test/Test_GeneyxPgxParser.py -v
pip install -e . compiles the modules in place, and Python loads a
.pyd/.so in preference to the .py beside it — so after an edit the tests
keep running the copy built at install time until you reinstall. The
symptom is a change that provably has no effect. Delete the build artefacts to
go back to running the sources:
find GeneyxPgxParser -name "*.pyd" -delete -o -name "*.so" -delete -o -name "*.c" -delete
They are gitignored, and CI builds the wheels from source regardless.
Test fixtures are real PacBio output for HG002, a public GIAB reference sample, ONT output for NA19207, and Thermo's own PharmacoScan 24F demo plate unmodified — 24 samples × 32 genes, two Thermo controls and 22 1000 Genomes / Coriell cell lines, as shipped with Axiom Analysis Suite. The whole plate is kept rather than a cut-down copy because its awkward calls are spread across samples: between them they cover every interpretation code the format has. No patient data belongs in this repository.
Metadata
Release files for GeneyxPgxParser 0.1.8
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distributions (wheels)
Total release size: 60.3 MB
Release files / geneyxpgxparser-0.1.8-cp314-cp314-win_amd64.whl
| Download URL | geneyxpgxparser-0.1.8-cp314-cp314-win_amd64.whl |
|---|---|
| Size | 1.6 MB |
| Tags | CPython 3.14 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
27bde3a93a335c24e6393fd3f5e50e66a2e55f7da055f221cdd9ed86b707437f
|
|
BLAKE2b-256 checksum How to use checksums |
bb8b70eeaf6ae4c6f77e7283ffd15a06d34a6c8a9ff93352c9a90f8143d64e46
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp314-cp314-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 6.8 MB |
| Tags | CPython 3.14 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
2448802ff34d654af761b47fc4e46c02580abfb8e2508635e69034d8a3b31343
|
|
BLAKE2b-256 checksum How to use checksums |
fb9aa355a3041cdba434607140a89974ea8c67a3e2af65d9ef60ed2fb7fe685c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp314-cp314-macosx_10_15_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp314-cp314-macosx_10_15_x86_64.whl |
|---|---|
| Size | 1.7 MB |
| Tags | CPython 3.14 macOS 10.15+ x86-64 |
|
SHA-256 checksum How to use checksums |
ce55fe163dbc89bf5e2dbe68439c36092e7de7195917f370e4aff93584d62eaf
|
|
BLAKE2b-256 checksum How to use checksums |
f4ca765c09dc3e8d0180a6f42c23a871b17382c26d55a7b175aff158f82d10f6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp313-cp313-win_amd64.whl
| Download URL | geneyxpgxparser-0.1.8-cp313-cp313-win_amd64.whl |
|---|---|
| Size | 1.6 MB |
| Tags | CPython 3.13 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
8f63d4d2a1cacfd7331f41edf77d9e835123a514d3dec2c4233c53e8053a12a3
|
|
BLAKE2b-256 checksum How to use checksums |
87dd08ee771aaf6e4fb47b969bf95b42e956f7fc0c88230f16858d5c121f649e
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp313-cp313-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 6.8 MB |
| Tags | CPython 3.13 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
2634f40fecc791e3f5142f0fe70430d7452a08fd8ea36fac23ad5afc6dfaa6af
|
|
BLAKE2b-256 checksum How to use checksums |
36a27b3f1a4585b7b86427a8be2acddcf095eb2b50f4955c5a05499d81266d04
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp313-cp313-macosx_10_13_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp313-cp313-macosx_10_13_x86_64.whl |
|---|---|
| Size | 1.7 MB |
| Tags | CPython 3.13 macOS 10.13+ x86-64 |
|
SHA-256 checksum How to use checksums |
3641d3e5a9c910b3df358888149292b90d05f9bf7408cfe0707ec6f3b5952bab
|
|
BLAKE2b-256 checksum How to use checksums |
a4f196c5c8ef9c349f18e892b9a4248e7fec736ce40778f864ced2059e0940f5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp312-cp312-win_amd64.whl
| Download URL | geneyxpgxparser-0.1.8-cp312-cp312-win_amd64.whl |
|---|---|
| Size | 1.6 MB |
| Tags | CPython 3.12 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
4dcd56392126f199436f991fcfe0a20f83bfb79b60a410bc053a11a79142f298
|
|
BLAKE2b-256 checksum How to use checksums |
fd9a9b6192277624a6e1ba72fe12c0a0f4efabb7b9bf061c9b26e1b4a20c9895
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 6.9 MB |
| Tags | CPython 3.12 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
521a14067e823a98011797068e093cbcc79491d98e07aadc1311512283767675
|
|
BLAKE2b-256 checksum How to use checksums |
24f4e2e849b915fd4af45a0420f590c7992fa4bbf23e82671cfb33bcceb040e5
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp312-cp312-macosx_10_13_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp312-cp312-macosx_10_13_x86_64.whl |
|---|---|
| Size | 1.7 MB |
| Tags | CPython 3.12 macOS 10.13+ x86-64 |
|
SHA-256 checksum How to use checksums |
650ee96746ca80f0ca53a324526d15d5fbcd849859b1b7e89644ddc1acd34414
|
|
BLAKE2b-256 checksum How to use checksums |
3e0661306aa480e581d7949e716a1b5dd7c124d730365da2ad0f766bbad28e8d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp311-cp311-win_amd64.whl
| Download URL | geneyxpgxparser-0.1.8-cp311-cp311-win_amd64.whl |
|---|---|
| Size | 1.6 MB |
| Tags | CPython 3.11 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
939040f496c8a2c88c71379be7e9709ab329841f2e93d1eae5a883bcb7c4040b
|
|
BLAKE2b-256 checksum How to use checksums |
1ac59763a9baf659e56f439ae60d2704e23f2c188f4a4594192cc03c685c062d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 6.9 MB |
| Tags | CPython 3.11 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
cc38baecd654590aec2e98c294a02be737f80aaf0cea924be1a970a84d9db753
|
|
BLAKE2b-256 checksum How to use checksums |
cb2477b12b127bcf7ae0fdcbcf0f0398012dbe0b528d56781430da0459ec25f2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp311-cp311-macosx_10_9_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp311-cp311-macosx_10_9_x86_64.whl |
|---|---|
| Size | 1.7 MB |
| Tags | CPython 3.11 macOS 10.9+ x86-64 |
|
SHA-256 checksum How to use checksums |
053579cc23fa555b830563b2dfe1833a4da55b7556bceb91dff61b76d109c74e
|
|
BLAKE2b-256 checksum How to use checksums |
6655181c0073f4a1fb0b1b67925b2333c2680323654e6cf8d3a92ae9494880d4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp310-cp310-win_amd64.whl
| Download URL | geneyxpgxparser-0.1.8-cp310-cp310-win_amd64.whl |
|---|---|
| Size | 1.6 MB |
| Tags | CPython 3.10 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
7f701c7d4d2c6ac1b0c4c639b0c9a184e0c5a63a5804c42870b69c67e86c4d70
|
|
BLAKE2b-256 checksum How to use checksums |
82071c0f1849e76fe8d434a22b50500caa79d26cef54a582932dc9dea6eb5ad6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 6.6 MB |
| Tags | CPython 3.10 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
6f7538b41c34c9576c01ce4ea94d976aa9670c8d8bab5647910fe44f07d516f1
|
|
BLAKE2b-256 checksum How to use checksums |
e4038ca771bf55fdd3110ea34ffd40ec9977ad4b7b399a0c98ef41845f3d63ae
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp310-cp310-macosx_10_9_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp310-cp310-macosx_10_9_x86_64.whl |
|---|---|
| Size | 1.7 MB |
| Tags | CPython 3.10 macOS 10.9+ x86-64 |
|
SHA-256 checksum How to use checksums |
ced92db43d4a68772711709588f633b857d4e05d882b6aaa244ad31d290cbd25
|
|
BLAKE2b-256 checksum How to use checksums |
12458c6585e721388834e9d20ea7fde4c5eabc11fbb9d7401af59c57953a02b9
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp39-cp39-win_amd64.whl
| Download URL | geneyxpgxparser-0.1.8-cp39-cp39-win_amd64.whl |
|---|---|
| Size | 1.6 MB |
| Tags | CPython 3.9 Windows x86-64 |
|
SHA-256 checksum How to use checksums |
7e41f5bd9e1d9057f44c0d4e64601ed984f79be30d6890ba9987e9ecc0cffb69
|
|
BLAKE2b-256 checksum How to use checksums |
9d83410fb63137bcdb74755a06ec7c1abe3c9f9f86cbd12cfcb75763ff07a8e1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.manylinux_2_28_x86_64.whl |
|---|---|
| Size | 6.6 MB |
| Tags | CPython 3.9 Linux glibc 2.17+ x86-64 Linux glibc 2.28+ x86-64 |
|
SHA-256 checksum How to use checksums |
f748d370ba56ea134fb9b7035760325bc7b4e98670a565f8a50e6f3600273c35
|
|
BLAKE2b-256 checksum How to use checksums |
6258d743f6b40e8eeae9078fd39073526a2b7d37ecb00fc6f1f29439286a9c1c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Release files / geneyxpgxparser-0.1.8-cp39-cp39-macosx_10_9_x86_64.whl
| Download URL | geneyxpgxparser-0.1.8-cp39-cp39-macosx_10_9_x86_64.whl |
|---|---|
| Size | 1.7 MB |
| Tags | CPython 3.9 macOS 10.9+ x86-64 |
|
SHA-256 checksum How to use checksums |
adf1977aaf358ae54dbc17960c426f1a48cbb6feb9435f98ecba789b37bac7c8
|
|
BLAKE2b-256 checksum How to use checksums |
5c973a9d57b909876cdb8608e4c740064e4c8b786f6aafbe269438a47acc5552
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.11.17
|
Security for GeneyxPgxParser 0.1.8
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.