inspect

Table of contents

1. Description
2. Usage
3. Haplotype output
4. Command line options
   1. Basic options
   2. Input files
   3. Output files

Description

Inspect a GLIMPSE2 binary reference panel (.bin file produced by GLIMPSE2_split_reference), print summary statistics and, optionally, write the reference haplotypes it contains as VCF/BCF. Useful for debugging, validation, and understanding chunk characteristics, e.g. to see exactly which sites and haplotypes GLIMPSE2_phase was given for a chunk.

Statistics reported include:

• File size, chromosome, input/output regions (bp and Mbp)
• Genetic map span for input and output regions (cM), and whether the cM positions were interpolated from a genetic map or use the constant 1 cM/Mb fallback
• Number of haplotypes
• Variant counts (total, common, rare, common HQ, low quality)
• Variant types (SNP, MNP, indel, other)
• Allele-frequency distribution (monomorphic, singletons, MAC 2-5, MAF <1%, 1-5%, 5-50%)
• Core (output region) vs buffer variant breakdown

Usage

GLIMPSE2_inspect --input reference_panel/split/1000GP.chr22.noNA12878_chr22_15649863_19744259.bin

To also write the haplotypes of the chunk’s core (output) region:

GLIMPSE2_inspect --input reference_panel/split/1000GP.chr22.noNA12878_chr22_15649863_19744259.bin --output chunk_01.bcf

Haplotype output

With --output, the reference haplotypes stored in the .bin file are written as VCF/BCF (format chosen from the file extension: .bcf, .vcf.gz or .vcf; compressed outputs are indexed with a .csi file).

• Samples: the .bin file stores neither sample names nor ploidy, so each haplotype is written as its own haploid sample, in the order of the original panel, named h1, h2, … zero-padded to a fixed width (e.g. h00001 to h18000 for 18,000 haplotypes). For an all-diploid panel, haplotypes 2i-1 and 2i are the two phased haplotypes of the i-th original sample.
• Sites: CHROM, POS, ID, REF and ALT as in the original panel. Only the sites GLIMPSE2_split_reference kept are present: multi-allelic records and (unless --keep-monomorphic-ref-sites was used) monomorphic records were dropped when the panel was split. QUAL, FILTER, other INFO and FORMAT fields (including INFO/END, so symbolic alleles lose their span), and unphased status are not stored.
• INFO fields: AC, AN and AF in the reference panel; RARE flags variants stored sparsely (minor allele frequency below --sparse-maf); CM is the genetic position relative to the first core variant of the chunk, not a position on the chromosome’s genetic map (variants in the left buffer are negative). CM is a 32-bit float: text VCF shows 6 significant digits, so nearby variants can show the same value. BUFFER flags buffer variants when --include-buffers is used.
• Region: by default only the core region is written; --include-buffers also writes the variants in the buffers on either side.

Command line options

Basic options

Input files

Output files