Sniffles2 Documentation

To call structural variants (SVs) from long-read alignments (PacBio or Oxford Nanopore), provide a BAM file as input and specify a VCF file for the output.

sniffles -i mapped_input.bam -v output.vcf

You can install Sniffles2 using either pip or conda. If you are upgrading from Sniffles1 installed via conda, use the update command.

# Using pip
pip install sniffles

# Using conda
conda install sniffles=2.8.0

# Upgrading from Sniffles1 via conda
conda update sniffles=2.8.0

To determine the genotype of a specific set of SVs from an existing VCF file for a given sample, use the --genotype-vcf option.

sniffles --input sample.bam --genotype-vcf input_known_svs.vcf --vcf output_genotypes.vcf

To detect mosaic structural variants (e.g., somatic SVs), add the --mosaic flag to your command.

sniffles --input mapped_input.bam --vcf output.vcf --mosaic

Multi-sample calling is a two-step process. First, generate .snf files for each individual sample. Second, combine those files into a single multi-sample VCF.

Step 1: Create .snf files for each sample sniffles --input <sample>.bam --snf <sample>.snf

Step 2: Combine files You can combine files by listing them individually or by providing a .tsv file containing a list of .snf files (and optional custom sample IDs in a second column).

# Step 1: Generate .snf files
sniffles --input sample1.bam --snf sample1.snf
sniffles --input sample2.bam --snf sample2.snf

# Step 2a: Combine using individual .snf files
sniffles --input sample1.snf sample2.snf --vcf multisample.vcf

# Step 2b: Combine using a .tsv list
sniffles --input snf_files_list.tsv --vcf multisample.vcf

When phasing is enabled in the configuration (config.phase), Sniffles2 modifies the VCF output in two ways:

1. Genotype Separator: The genotype separator changes from / (unphased) to | (phased) if the haplotype information is available.
2. PS Tag: The PS (Phase-block) tag is added to the FORMAT column.

Phasing logic uses hp_i (the index of the haplotype in config.phase_identifiers) to determine the orientation of the phased genotype (e.g., 1|0 vs 0|1).

Mosaic mode is activated if mosaic_include_germline is set. When in mosaic mode, Sniffles2 adjusts several QC and clustering parameters:

• qc_nm_measure: Uses mosaic_qc_nm if the standard qc_nm is not provided.
• cluster_merge_len: Uses default_cluster_merge_len_mosaic if the current cluster_merge_len matches the standard default_cluster_merge_len.

Sniffles2 applies different coverage and support calculation strategies depending on the type of Structural Variant (SV) to ensure accuracy:

Multi-sample calling is a two-step process:

1. Generate SNF files: Run Sniffles2 for each individual sample using the --snf flag to produce a .snf file containing SV candidates.
2. Combined calling: Run Sniffles2 using the generated .snf files as input to produce a combined VCF.

You can provide multiple .snf files directly or use a .tsv file containing a list of .snf files (one per line).

# Step 1: Create .snf for each sample
sniffles --input sample1.bam --snf sample1.snf
sniffles --input sample2.bam --snf sample2.snf

# Step 2: Combined calling
sniffles --input sample1.snf sample2.snf --vcf multisample.vcf

When genotyping known structural variants, you can configure the output format and ploidy.

• genotype_format: The VCF format string used for genotype information. The default is "GT:GQ:DR:DV".
• genotype_ploidy: Specifies the ploidy of the sample. Note: Currently only --genotype-ploidy 2 is supported.
• genotype_min_z_score: The minimum Z-score required for genotyping.

# Example genotyping command
sniffles --genotype-ploidy 2 --genotype-format "GT:GQ:DR:DV" ...

Consensus calling is governed by several parameters that define how k-mers are used to validate variants:

• consensus_min_reads: Minimum reads required for consensus (default: 4).
• consensus_kmer_len: Length of the k-mer (default: 6).
• consensus_kmer_skip_base: Number of bases to skip between k-mers (default: 3).
• consensus_kmer_skip_seqlen_mult: Multiplier for skipping based on sequence length (default: 1.0 / 500.0).
• consensus_low_threshold: Threshold for low-confidence consensus (default: 0.0).