pipelines-docs

mSNV Calling and Filtering Pipeline — SmahtSNV 2.0.0

Updates in SmahtSNV 2.0.0

The following changes have been introduced relative to the previous pipeline version (v1.0.0):

  1. Core-specific variant tracking: Each GCC sequencing core is tracked independently throughout the pipeline. The final VCF is multi-sample, with one column per core recording per-core read counts, variant allele fractions, and caller information.

  2. CrossCore evidence label: A fourth cross-evidence flag has been added. CrossCore is set when a variant is independently detected and supported in 2 or more sequencing cores from the same tissue (see Cross-Evidence Classification). It complements the existing CrossTech, CrossCaller, and CrossTissue labels.

  3. Revised confidence system — PASS / LowEvidence / EvidenceScore: The previous three-class system (HighConf / LowConf / LikelyArtifact) has been replaced. EvidenceScore is an integer INFO field (0–4) counting how many Cross* flags are present. Variants with at least one Cross* flag receive a FILTER of PASS (EvidenceScore ≥ 1); variants with no corroborating evidence receive LowEvidence (EvidenceScore = 0). Downstream analyses can apply additional stratification using EvidenceScore directly.

  4. Updated CrossTissue reporting: Tissues from the same donor with short-read support for a variant are now listed in the SR_TISSUE_PRESENCE INFO field (replaces the previous TISSUE_SR_VAFs field).

Overview

The SMaHT mosaic SNV pipeline (SmahtSNV 2.0.0) integrates four somatic SNV callers — three short-read-based and one long-read-based — followed by hierarchical filtering and multi-axis cross-evidence validation to generate high-confidence mosaic SNV calls. Both merged sequencing libraries from multiple genome centers (GCCs) as well as core-specific libraries for each donor are provided as high-depth input (~300× combined, or ~150x core-specific short-read coverage) to each caller. Starting in v2.0.0, each sequencing core is tracked independently, and the final output is a multi-sample VCF with one column per core.

Key Pipeline Steps

Calling → Short-read filtering → Long-read validation → Cross-evidence tagging → Confidence assignment

Home - Overview - Somatic SNV Callers - Short-Read Based Filtering - Long-Read Based Filtering - Cross-Evidence Classification - Confidence Designation