GLIMPSE (Genotype Likelihoods IMputation and PhaSing mEthod)

GLIMPSE2 is a suite of tools designed for low-coverage whole-genome sequencing imputation, optimized for large reference panels (hundreds of thousands of samples) and rare variants. The suite consists of the following core tools:

• chunk: Used to phase common sites (e.g., SNP array data) or as the initial step for WES/WGS data.
• split_reference: Used to phase common sites, typically SNP array data, or as the first step for WES/WGS data.
• phase: Ligate multiple phased BCF/VCF files into a single whole chromosome file. This is typically used to ligate multiple chunks of phased common variants.
• ligate: Phases rare variants onto a scaffold of common variants (the output of phase or ligate).
• concordance: A program to compute switch error rates and genotyping error rates using simulated or trio data.

GLIMPSE (Genotype Likelihoods IMputation and PhaSing mEthod) is a method designed for haplotype phasing and genotype imputation of low-coverage Whole Genome Sequencing (WGS) data.

There are two primary versions available:

• GLIMPSE2: The latest version, optimized for high accuracy and computational performance, particularly when using large reference panels for rare variants.
• GLIMPSE1: The original version of the tool.

Developers and researchers can access documentation and research papers via the official GLIMPSE websites.

GLIMPSE2 is a C++ toolset for haplotype phasing in large datasets. It is compatible with the following architectures and compilers:

• Linux x86_64: Requires GCC or Clang.
• Linux aarch64/ARM64: Requires GCC or Clang.
• macOS ARM64 (Apple Silicon): Requires Clang (via Xcode Command Line Tools).

In the context of genetic variation, a Singleton is a rare variant where the minor allele count is exactly 1. This means the genetic variation is carried by a unique chromosome within the dataset.

Because these variants are unique, there is no existing information to phase them using standard statistical population-based methods, which typically result in a 50% switch error rate. While family information is the traditional way to phase singletons, SHAPEIT5 can provide non-random phasing without family data by utilizing a coalescent-inspired model.

Low-coverage whole genome sequencing (WGS) refers to sequencing data with low depth, typically ranging between 0.1x and 8x coverage.

Because coverage is low, genotype calling is often uncertain. To make this data usable for downstream processing, it frequently requires imputation using specialized methods. GLIMPSE is a specialized tool designed for the imputation of low-coverage WGS data.

Haplotype phasing is the process of distinguishing the two parentally inherited copies of each chromosome into individual haplotypes.

SHAPEIT5 is a specialized tool designed for statistical population-based haplotype phasing. It is specifically optimized to focus on extremely rare variants.

In the context of GLIMPSE, a reference panel is a large set of deeply-sequenced haplotypes. These panels are used for two primary genomic tasks:

1. Genotype Imputation: Predicting missing genotypes in a sample.
2. Haplotype Phasing: Determining the arrangement of alleles on individual chromosomes.

Reference panels are typically used for cohorts that have been genotyped using SNP arrays or low-coverage Whole Genome Sequencing (WGS).