Research small-variant evidence report

CHANGED_ASSAY_MGI

MutationSieve 0.1.0.dev1 · somatic mode · sample state VARIANT_CANDIDATE_PRESENT_PARTIAL_NO_CALL

Not analytically or clinically validated. A REVIEW record is a sequencing hypothesis, not a diagnosis. A missing record is not a clinical negative. HIGH/PASS are disabled unless a separately validated, compatible analytical resource set is explicitly enabled.

Plain-language summary

The program examined high-quality DNA sequencing evidence inside the declared target regions. It found 4 candidate(s) worth human review and retained 0 additional candidate(s) with reasons they were rejected. It could assess approximately 100.0% of the target bases at the configured depth. These numbers describe this software run; they do not establish whether a person does or does not have a mutation.

4non-rejected hypotheses

0audited rejects

100.0%basewise primary-unit callable target

4,181alignment records used

0.006%high-quality reference mismatch fraction

absenterror model status

Candidate evidence

Alternate and reference depth use the same primary unit. Coordinate-family counts are shown only as an audit proxy and are not called molecules. QUAL is deliberately missing in VCF because the rank is not an independently calibrated probability.

Tier	Allele	Support	Strands	Haplotype	Statistics	Rank	Filters and review flags
REVIEW REVIEW_CANDIDATE	`chrSynthetic:1201 T>A` SNV · GENE_A	14/41 34.146% · paired_fragment	5/9 forward/reverse	14 local-haplotype favored	8.76e-38 q 1.83e-30	0.989 rank, not probability	review: empirical_resolution_insufficient_for_reporting; review: exact_locus_unassessable_in_panel_of_normals; review: no_compatible_empirical_error_model; review: no_matched_normal_tumor_only_analysis; review: no_validated_umi_family_consensus
REVIEW REVIEW_CANDIDATE	`chrSynthetic:2201 C>A` SNV · GENE_A	8/54 14.815% · paired_fragment	3/5 forward/reverse	8 local-haplotype favored	6.46e-19 q 6.75e-12	0.983 rank, not probability	review: empirical_resolution_insufficient_for_reporting; review: exact_locus_unassessable_in_panel_of_normals; review: no_compatible_empirical_error_model; review: no_matched_normal_tumor_only_analysis; review: no_validated_umi_family_consensus
REVIEW REVIEW_CANDIDATE	`chrSynthetic:4480 C>CAA` INS · GENE_H	5/45 11.111% · paired_fragment	3/2 forward/reverse	3 local-haplotype favored	1 q 1	0.057 rank, not probability	review: conflicting_or_multiallelic_fragment_evidence; review: empirical_resolution_insufficient_for_reporting; review: exact_locus_unassessable_in_panel_of_normals; review: no_compatible_empirical_error_model; review: no_matched_normal_tumor_only_analysis; review: no_validated_umi_family_consensus; review: repeat_or_homopolymer_indel_context
REVIEW REVIEW_CANDIDATE	`chrSynthetic:5299 CAGA>C` DEL · GENE_H	4/47 8.511% · paired_fragment	2/2 forward/reverse	4 local-haplotype favored	1 q 1	0.138 rank, not probability	review: empirical_resolution_insufficient_for_reporting; review: exact_locus_unassessable_in_panel_of_normals; review: no_compatible_empirical_error_model; review: no_matched_normal_tumor_only_analysis; review: no_validated_umi_family_consensus

What the pipeline did

Validated the BGZF/BAM container, BAM records, reference identity, contig lengths, and analysis configuration.
Generated permissive SNV and small-indel candidates from quality-qualified aligned bases and CIGAR events.
Re-read every candidate locus, reconciled overlapping mates, separated fragments, coordinate-family audit groups, and explicit UMI families, and rescored supporting reads against local reference and alternate haplotypes.
Compared evidence with a compatible panel of normals when one was supplied, adjusted over the declared full callable search rather than emitted candidates alone, and applied explicit hard and soft rules.
Wrote normalized VCF, reviewer and complete-audit TSVs, JSON, QC, callability, HTML, and a hash-validated manifest.

Important limitations

Research software only; no assay-specific analytical or clinical sensitivity, specificity, PPV, NPV, LoB, LoD95, precision, reproducibility, robustness, or clinical validity has been established.
Candidate discovery is CIGAR-led. Local haplotype rescoring checks an emitted allele but is not de-novo assembly, so misaligned or unrepresented indels can be missed.
The built-in BAM path does not read CRAM. CRAM requires the optional HTSlib/pysam backend and exact reference; FASTQ requires a separately validated aligner.
Flagged duplicates are excluded by default. Coordinate-family grouping is not a substitute for a quality-aware, assay-validated UMI/duplex consensus workflow.
Target depth outside the declared bundle, large indels, complex structural variants, copy-number changes, fusions, repeat expansions, and RNA variants are out of scope.
Tumor-only analysis cannot reliably distinguish somatic from germline variation. Matched normal evidence is required before somatic HIGH/PASS can be enabled.
No transcript consequence, population frequency, oncogenicity, pathogenicity, or treatment interpretation is performed.

How to use the files

variants.vcf: Normalized non-rejected hypotheses for standard genomics tools; genotype and QUAL remain missing.
variants.tsv: Human-sortable review list.
all-candidates.tsv and mutationsieve.json: Complete audit trail, including rejected candidates and exact reasons.
callability.bed and qc.json: Basewise depth plus maximum same-unit REF span at each start; MNV/indel negatives require enough span, and incomplete territory cannot support a negative interpretation.
manifest.json: Completion marker and SHA-256 identity of the other output files.

Resolved configuration

{
  "assay": "somatic",
  "assay_compatibility_token": "",
  "cram_reference_fasta": "",
  "drop_duplicates": true,
  "enable_reportable_tiers": false,
  "fdr": 0.01,
  "full_search_alleles_per_base": 2612,
  "llm_band": [
    0.25,
    0.9
  ],
  "llm_enabled": false,
  "llm_max_candidates": 25,
  "llm_max_logit_shift": 1.5,
  "llm_model": "claude-opus-5",
  "max_candidates": 25000,
  "max_events_per_record": 20000,
  "max_indel_length": 50,
  "max_mnv_length": 10,
  "max_normal_af": 0.02,
  "max_pending_mates": 50000,
  "max_query_bases_per_record": 1000,
  "max_records": 1000000,
  "max_recount_observations": 1000000,
  "max_reference_mismatch_fraction": 0.08,
  "max_soft_clip_fraction": 0.25,
  "max_umi_callability_observations": 1000000,
  "min_af_germline": 0.2,
  "min_af_mosaic": 0.005,
  "min_af_somatic": 0.01,
  "min_alt_mean_base_quality": 25.0,
  "min_alt_mean_mapq": 30.0,
  "min_alt_start_positions": 1,
  "min_alt_support": 1,
  "min_base_quality": 20,
  "min_depth": 3,
  "min_family_consensus": 0.9,
  "min_family_members": 2,
  "min_mapq": 20,
  "min_normal_depth": 1,
  "min_raw_alt_reads": 1,
  "min_umi_tag_fraction": 0.95,
  "normal_bam": "",
  "outdir": "${MUTATIONSIEVE_PROJECT_ROOT}\\benchmark\\sample-results\\CHANGED_ASSAY_MGI",
  "pon_max_sample_fraction": 0.5,
  "pon_min_alt_support": 2,
  "pon_min_samples": 2,
  "pon_path": "",
  "read_end_fraction": 0.1,
  "reference_dir": "${MUTATIONSIEVE_PROJECT_ROOT}\\benchmark\\reference",
  "require_umi_consensus": false,
  "sample": "CHANGED_ASSAY_MGI",
  "strand_bias_p": 0.001,
  "threads": 1,
  "tier_high": 0.9,
  "tier_pass": 0.65,
  "tier_review": 0.25,
  "umi_tags": [
    "RX",
    "MI"
  ]
}

Provenance and compatibility identity

{
  "analysis_fingerprint": "8c8984896a2f39c160afd782cf1be2bf9470d0140f2056760245b261f74e0040",
  "inputs": {
    "case": {
      "path": "${MUTATIONSIEVE_PROJECT_ROOT}\\benchmark\\synthetic-inputs\\CHANGED_ASSAY_MGI.bam",
      "sha256": "7dba8072720aedcad175cced5ebc406e46e510e443954443f29ae3e3b8cda4b0",
      "size": 85974
    }
  },
  "methods": {
    "candidate_discovery": "quality_filtered_bases_and_cigar_events",
    "coordinate_family_use": "conservative_audit_proxy_not_molecules",
    "indel_reconciliation": "local_reference_vs_alternate_haplotype_rescore;_not_denovo_assembly",
    "llm_effect_on_calling": "none",
    "multiplicity": "benjamini_hochberg_complete_declared_search_family",
    "primary_counting_unit": "paired_fragment_or_explicit_umi_consensus"
  },
  "pon_compatibility": {
    "assay_compatibility_token": "",
    "assay_mode": "somatic",
    "assembly": "synthetic-MutationSieve-v1",
    "drop_duplicates": true,
    "max_events_per_record": 20000,
    "max_indel_length": 50,
    "max_mnv_length": 10,
    "max_query_bases_per_record": 1000,
    "max_records": 1000000,
    "max_recount_observations": 1000000,
    "max_umi_callability_observations": 1000000,
    "min_base_quality": 20,
    "min_depth": 3,
    "min_family_consensus": 0.9,
    "min_family_members": 2,
    "min_mapq": 20,
    "min_raw_alt_reads": 1,
    "min_umi_tag_fraction": 0.95,
    "primary_analysis_unit": "paired_fragment",
    "reference_manifest_sha256": "e92e221ac743171eb4ae4fee44e9a61c2110998038c38bc9aa2e6f83da4cc658",
    "require_umi_consensus": false,
    "schema": "mutationsieve.pon-compatibility.v1",
    "sequencing_platforms": [
      "MGI"
    ],
    "target_coordinate_system": "0-based-half-open",
    "target_hash": "36fbb4dea6a8ded616d8a3f1a3017ae1d4f5e3e8fa6ca17e9c6d00ca0851b89a",
    "umi_tags": [
      "RX",
      "MI"
    ]
  },
  "reference": {
    "files": [
      {
        "path": ".benchmark-source.fa",
        "sha256": "92ce866a52012e4e8b748c8f954dccc4b052e6ce938c6b59b4bdea09574c469b",
        "size": 12214
      },
      {
        "path": ".benchmark-source.fa.fai",
        "sha256": "f16fbc938221d1d047b3dc9f98ddd323c378f8c643c4adaf53b2ad0ce1bdd7db",
        "size": 28
      },
      {
        "path": ".benchmark-targets.bed",
        "sha256": "90aa19e3f6c2e4ca432ee06f79d3338e7ca2824c33b61fa105d2304107ed92b2",
        "size": 144
      },
      {
        "path": "panel.fa",
        "sha256": "e5e631f6a91d07c585fe6400f19022f750da5f67b157836e0357e154211fc382",
        "size": 9974
      },
      {
        "path": "panel.json",
        "sha256": "e92e221ac743171eb4ae4fee44e9a61c2110998038c38bc9aa2e6f83da4cc658",
        "size": 2697
      }
    ],
    "root": "${MUTATIONSIEVE_PROJECT_ROOT}\\benchmark\\reference"
  },
  "schema": "mutationsieve.provenance.v1",
  "source": {
    "__init__.py": "8232c3610352162267aaeeab155ba8deb4ef210d9ba1beb8c82407a9b3df3893",
    "__main__.py": "3d2438ae92f4746e9ffaa1f837b1fe803f6281091d1f31f8c95643716f257d9a",
    "_spikeforge_policy.py": "5f5374a747036d8dd827de75bbb31fb2c5b71a0e22d092f87cddcd022de5cb43",
    "_spikeforge_worker.py": "34f4101af4a63b5dd80ff71a2ac007ffd90720a14307e5338835e6a0df824968",
    "alignment.py": "1efa4b0b78323ae94dc2f23138b02b0151f1e364924bf3d95c4c8a646dee4722",
    "bamio.py": "475aa5f007affbd037a315aa1b2a83b917933fc7e0c64417e279852987aebe41",
    "batch.py": "2cc4befa2ec61e18fecaa83c0bbe1e06d8618052d35300d0d03e0e6bd3a94a68",
    "benchmark.py": "bc73aaf53790a4f61eac37d718e91afd3bdfe3b616e8f65dfe24a4f56343a16d",
    "cli.py": "c76f2e8b7baa83cb9c50f7f09bf97b25c894f18b959c7c79cad2cdf1d564b89c",
    "config.py": "1460d0622d4730ab2d6abd67fc6e7ee6a27cd430480c5f2befdf448f798ad35e",
    "evidence.py": "ab789ed953e7a7266d4dac60d88fa79d50b314defcabb30fbed42df8eb0d9349",
    "external_validation.py": "3f786e9ee88887e7576e87f5fddbbf47259a10c8e5a1a3eca6837eba6e2e84db",
    "llm.py": "ab6aaea8de8794b801fdff6edd973dec73b8a8dabc3f572067ee4db2475e1c7b",
    "outputs.py": "1db4f97d6e69ee6989422f8c5eefbabc8198a4e30b5fb9e3a1b52eb6976257eb",
    "pipeline.py": "e9ce2192bd7cb897cb8a122651eb9161e8178c5d73a6a27963afb55a6d460334",
    "pon.py": "acf08f6dec233aa02bd1e21ff74342b44b7e117171168d963370021626f652bc",
    "provenance.py": "aeed1b20076c01a8457ff6d88c4f97ed373a2a53c0ecc8106728b3b9180f9cd1",
    "reference.py": "6a6f43ac2ca5e6f9eeb0ec34bbc6446e0a3007b87df5afbd03eebe75ff5856d2",
    "release.py": "ab70ee440a215cd151d770824b0fe1d91bc8c1db6af2795b219a28f78c9f822c",
    "report.py": "b9541ea6662e4f28d5847c8ef9705ddb278a00d5d2a8052ab2923d5120d85717",
    "representation.py": "0f920799e7078c98e075bebd34ecd043df7773555666946932c618ab2138e094",
    "rules.py": "1d847f8ce8754ae1b8e912888d4c7eafea37234050c8734d9661778dd77ccadf",
    "simulate.py": "e0200f42fbb76b7cab8097548524243a6ad43598023f131e138ac4b3535d5612",
    "stats.py": "d4b6dbb6f0f6666a9e7371f98ed700e4a7640230a435c68452c0f87f91de57ae",
    "validation.py": "e5867edc99b69cdb06b8bbfef368676087a0a50844d06e63b606c7e94a868792",
    "variants.py": "429acaa447b049c3ddf52ec42960c28a2eb3ced6d3556aa220152ef24703248d"
  }
}

Research use only. MutationSieve has not established assay-specific analytical or clinical sensitivity, specificity, PPV, NPV, limit of blank, LoD95, precision, reproducibility, clinical validity, or treatment relevance. A missing call is not a clinical negative and a candidate is not a diagnosis.
Analysis fingerprint: 8c8984896a2f39c160afd782cf1be2bf9470d0140f2056760245b261f74e0040