CHM13 Opportunities and Downstream Compatibility
What Is Included
MatrixGenerator supplies raw CHM13 SBS, DBS, and indel matrices, plus 48 opportunity files covering whole genome and annotation-derived CDS exome regions. The opportunity contexts are 6, 24, 96, 384, 1536, 6144, DBS, and DBS186.
Counts record the number of eligible reference windows per context/chromosome. Distribution rows divide each chromosome count by the total for that context across the selected chromosomes. These are chromosome-selection probabilities, not a mutation-signature catalogue and not a frequency distribution over contexts. Male tables include X/Y and female tables omit Y, following the existing resource convention; they do not model copy number or double the number of X copies. All CHM13 tables are nuclear-only.
SBS contexts use eligible five-base windows, even when reporting one- or three-base contexts. Unknown DNA bases invalidate a window. DBS uses two-base windows. Exome windows must fit within the CDS union, without padding or windows crossing excluded bases. Zero-opportunity rows have zero probabilities. Different capture kits or custom BED regions require their own opportunity counts; the shipped exome tables must not be substituted for arbitrary target regions.
See sources and build command and the runnable matrix example.
Extractor and Assignment Are Not Automatically Enabled
The reviewed local Extractor code accepts CHM13 matrices for de novo extraction,
but its catalogue-selection code in SigProfilerExtractor/sigpro.py permits
only GRCh37, GRCh38, mm9, mm10, mm39, rn6, and rn7. Other opportunity genome
names, including CHM13, are reset to GRCh37 with a console and metadata message
before the optional assignment/decomposition stage. Adding MatrixGenerator
resources does not change that list or supply CHM13 reference signatures.
Do not interpret a GRCh37 fallback as CHM13-adjusted catalogue matching. No downstream compatibility changes are included in this MatrixGenerator branch. Package versions may differ; inspect the installed downstream version and its metadata before interpreting results. Matrix counts themselves require no opportunity normalization and remain usable independently of catalogue matching.
Required Downstream Follow-Up
- Add an explicit CHM13 policy in Extractor/Assignment: permit de novo extraction
while rejecting unsupported catalogue matching rather than falling back to
GRCh37 without the user's explicit agreement. Test matrix and VCF entry points,
including
stop_after_extraction=True, and record both genome identities. - Define and validate how a chosen reference signature catalogue is adjusted from its original assembly to CHM13 opportunities. Specify the source assembly, whole-genome versus target-region opportunities, channel order, zero-count behavior, and normalization. Strand-aware/HD categories also need matching transcript/strand definitions; sequence ratios alone do not establish that.
- Register CHM13 only after the adjusted catalogue and regression expectations exist. Test decomposition, refitting, reported metadata, and plots. A new catalogue must not be presented as an official COSMIC CHM13 release.
- Audit Simulator's supported-genome lists and resource loading, then test real CHM13 WGS/exome simulations and female Y exclusion. File availability alone is not evidence that a downstream consumer supports the assembly.
These are separate downstream changes and review gates, not features already implemented by this branch. The packaged opportunity tables are a prerequisite.