STRC Research

STRC-STRCP1 Locus Identity Gate

4 min read

STRC-STRCP1 Locus Identity Gate

Rule

A variant score does not locate a DNA molecule. Assign an STRC-family haplotype to STRC or STRCP1 only when locus-specific evidence agrees. If the evidence is missing or conflicts, report unresolved and escalate; do not guess from the variant’s pathogenicity, phenotype fit, or the majority of nearby bases.

Computational gate

Run Paraphase on whole-genome PacBio HiFi reads aligned to the exact supported GRCh38 no-alt reference. Evaluate the configured pivot base together with the 314 bp STRCP1 deletion:

Pivot evidenceDeletion evidencePublic result
STRC/referenceabsentSTRC
STRCP1/alternatepresentSTRCP1
discordant pairanyunresolved (unknownhap)
missing, truncated, mixed, or inconsistentanyunresolved (unknownhap)

When any haplotype is unresolved, preserve independently supported total copy number but set gene-specific copy number to null. Omit the unresolved haplotype from the default locus-specific VCF. unknownhap is an abstention, not a claim of gene conversion or a hybrid allele.

This behavior is implemented in PacificBiosciences/paraphase#84. The independent test-only regression is PacificBiosciences/paraphase#83.

Escalation gate

An unresolved computational result returns to the laboratory with a concrete question: which physical molecule carries the disputed allele?

  1. Obtain targeted long reads from a gene-specific long-range PCR product whose primer or read span reaches sequence that distinguishes STRC from STRCP1.
  2. Require a single molecule or phased read set to bridge that locus anchor to the disputed variant. If no such bridge exists, locus identity stays unresolved.
  3. If copy number is uncertain, confirm it independently with validated MLPA or allele-specific ddPCR probes. Copy number corroborates the structure; it does not by itself place a small variant on STRC.
  4. Use parental samples when available to phase the locus structure and test the proposed inheritance model.

Targeted long-range PCR plus nanopore sequencing has distinguished STRC from STRCP1 in unresolved hearing-loss cases and can characterize suspected conversion structures. MLPA and allele-specific ddPCR provide orthogonal STRC copy-number evidence. See the targeted long-read study, the conversion study, and the ddPCR validation study.

Acceptance gate

A scientist can say STRC or STRCP1 only when the evidence chain contains:

  • exact reference and tool commit;
  • input provenance and checksums;
  • total and locus-specific copy-number evidence;
  • concordant locus markers or a physical molecule bridging a unique anchor to the disputed allele;
  • reproducible output; and
  • an explicit unresolved state for every unsupported assignment.

Pathogenicity classification starts only after locus identity. AlphaMissense, CADD, REVEL, phenotype fit, and ACMG evidence may help interpret a variant, but none can establish whether the assayed DNA molecule came from STRC or STRCP1.

Current evidence boundary

The public HG00733 regression checks same-input Python/Rust compatibility. The public HG005 gate independently checks a 3 STRC + 1 STRCP1 structure supported by MLPA and divergent markers in a non-peer-reviewed benchmark. Together they support conservative locus assignment and copy-number behavior; they are not a clinical validation study.

Downstream open-source domino

  1. Clinical-Genomics/paraphrase: normalize Paraphase v4 region_specific_info, then add explicit null/missing rule semantics without inventing a downstream-incompatible clinical status.
  2. Clinical-Genomics/scout: consume that identity state in the Dark Regions view and make unresolved STRC calls visually explicit before variant interpretation.
  3. nf-core/modules: after a Paraphase release containing the classifier, update the module and add a semantic STRC regression rather than only a generic smoke test.
  4. genomic-medicine-sweden/nallo: carry the same abstention through family aggregation in the active long-read rare-disease pipeline.
  5. nf-core/pacvar: after the module contract stabilizes, add Paraphase as the paralog-aware lane in the broader PacBio WGS workflow.

The immediate independent contribution is Paraphrase. The pipeline upgrades wait for the final Paraphase schema and a tagged release. PacBio Paraviewer also needs a direct safety fix because it can substitute total family CN when gene CN is explicitly unknown, but its proprietary license makes it source-available rather than open source.

Current implementation

Clinical-Genomics/paraphrase#16 implements the first domino: Paraphase 4 locus-specific fields are expanded into the existing flat handler, skip-key, and rule contract. The final head passed 20 tests and all lint hooks, including v3/v4 equivalence checks for CFH, NCF1, and STRC plus fail-fast mixed-schema cases.

The explicit null-versus-missing rule API remains on the local wip/paraphase-v4-call-state branch. It was deliberately separated because Scout currently accepts only normal, intermediate, and pathological and directly indexes that vocabulary. Locus callability must be coordinated as an identity/manual-review signal rather than smuggled into pathogenicity.

Connections

Graph View

Backlinks