GIGI: Genotype Imputation Given Inheritance

GIGI

Genotype Imputation Given Inheritance

Introduction

GIGI is a computer program to impute missing genotypes on pedigrees. Our approach handles large pedigrees by using a Markov Chain Monte Carlo-based program to infer inheritance vectors. Our approach has 2 steps:

1. use the program gl_auto from the MORGAN package to infer inference vectors using genotypes from informative framework markers

2. use GIGI to impute missing genotypes

Reference

Cheung, CYK., Thompson, E.A., Wijsman, E.M. (2013). GIGI: An approach to effective imputation of dense genotypes on large pedigrees. American Journal of Human Genetics 92:504-516 [link]

Download

The latest version of GIGI, along with example files, is available here (current version: 1.06.1).      [documentation] [FAQs] [changelog].
The program is written in C++ and runs on Linux machines.

Older version: v1.05

Older version: v1.04 [documentation]

Older version: v1.03

Older version: v1.02

Dependencies

This program depends on gl_auto output from MORGAN version 3.2. You can download MORGAN version 3.2 and find links to a user tutorial here.

We no longer support MORGAN version 3.1.1, but the gl_auto version 3.1.1 tutorial is available here for users who have files in this older format.

Related work

Blue EM, Cheung CYK, Glazner CG, Conomos MP, Lewis SM, Sverdlov S, Thornton T, Wijsman EM (2014) Identity-by-descent graphs offer a flexible framework for imputation and both linkage and association analyses. BMC Proceedings 8(Suppl 1):S19.

  • This GAW 18 article (describing a GAW project from 2012) is the first GIGI-related work that
    1. explores the idea of combining information from pedigree using GIGI and information from Linkage Disequibilirum using BEAGLE to impute genotypes.
      1. For imputation of rare alleles, the use of GIGI alone perfoms similarly to the combined approach, because effective imputation of rare alleles mainly comes from the pedigree with Inheritance Vectors.
      2. For imputation of common variants, our results suggest that combining BEAGLE with GIGI further improves imputation.
    2. gives support to the idea that the use of estimated probabilities to compute "dosage" to test for association between phenotype and SNPs may give higher power than the use of imputed results from genotype calls because "dosage" captures the uncertainties in imputated results.

    Note

    GIGI is for imputing genotypes on pedigrees with known pedigree structure. If the goal is to impute genotypes on unrelated individuals, consider using a population-based genotype imputation program (eg. BEAGLE) instead.
     

    Contact

    Method developer & maintainer: Charles Y K Cheung - cykc@uw.edu

    This work was supervised by Professor Elizabeth A. Thompson and Professor Ellen M. Wijsman

    last updated: February 2, 2015