GEMINI: a flexible framework for exploring genome variation



GEMINI (GEnome MINIng) is a flexible framework for exploring genetic variation in the context of the wealth of genome annotations available for the human genome. By placing genetic variants, sample phenotypes and genotypes, as well as genome annotations into an integrated database framework, GEMINI provides a simple, flexible, and powerful system for exploring genetic variation for disease and population genetics.

Using the GEMINI framework begins by loading a VCF file (and an optional PED file) into a database. Each variant is automatically annotated by comparing it to several genome annotations from source such as ENCODE tracks, UCSC tracks, OMIM, dbSNP, KEGG, and HPRD. All of this information is stored in portable SQLite database that allows one to explore and interpret both coding and non-coding variation using “off-the-shelf” tools or an enhanced SQL engine.

Please also see the original manuscript.


  1. GEMINI solely supports human genetic variation mapped to build 37 (aka hg19) of the human genome.
  2. GEMINI is very strict about adherence to VCF format 4.1.
  3. For best performance, load and query GEMINI databases on the fastest hard drive to which you have access.


In addition to the documentation, please review the following tutorials if you are new to GEMINI. We recommend that you follow these tutorials in order, as they introduce concepts that build upon one another.

  • Introduction to GEMINI, basic variant querying and data exploration. html pdf
  • Identifying de novo mutations underlying Mendelian disease html pdf
  • Identifying autosomal recessive variants underlying Mendelian disease html pdf
  • Identifying autosomal dominant variants underlying Mendelian disease html pdf
  • Other GEMINI tools html pdf

Latest news

Changes to Inheritance Tools

As of version 0.16.0, the built-in Mendelian inheritance tools are more stringent by default (they can be relaxed with the --lenient) option. By default, samples with unknown phenotype will not affect the tools, and strict requirements are placed on family structure. See the the docs for more info. In addition, the inheritance tools now support multi-generational pedigrees.

Much faster genotype queries

As of version 0.15.0, GEMINI creates auxilliary index files using the bcolz library. This allows queries that filter on sample genotype information via the --gt-filter option to be up to 1000 times faster than identical queries using version 0.14.0 or earlier. Details of the implementation and caveats can be found here.


In order to expedite your queries with the new bcolz indexing strategy, one must invoke the --use-bcolz option. We will likely phase this option out over time. For example:

gemini query \
  -q "select variant_id, gts.kid from variants"  \
  --gt-filter " == HET and == HET and gts.kid != 'HET'" \
  --use-bcolz \

New GEMINI Workflow

As version 0.12.2 of GEMINI it is required that your input VCF file undergo additional preprocessing such that multi-allelic variants are decomposed and normalized using the vt toolset from the Abecasis lab. Note that we have also decomposed and normalized all of the VCF-based annotation files (e.g., ExAC, dbSNP, ClinVar, etc.) so that variants and alleles are properly annotated and we minimize false negative and false positive annotations. For a great discussion of why this is necessary, please read this blog post from Eric Minikel in Daniel MacArthur’s lab.

Essentially, VCF preprocessing for GEMINI now boils down to the following steps.

  1. If working with GATK VCFs, you need to correct the AD INFO tag definition to play nicely with vt.
  2. Decompose the original VCF such that variants with multiple alleles are expanded into distinct variant records; one record for each REF/ALT combination.
  3. Normalize the decomposed VCF so that variants are left aligned and represented using the most parsimonious alleles.
  4. Annotate with VEP or snpEff.
  5. bgzip and tabix.

A workflow for the above steps is given below.

# setup

# decompose, normalize and annotate VCF with snpEff.
# NOTE: can also swap snpEff with VEP
#NOTE: -classic and -formatEff flags needed with snpEff >= v4.1
zless $VCF \
   | sed 's/ID=AD,Number=./ID=AD,Number=R/' \
   | vt decompose -s - \
   | vt normalize -r $REF - \
   | java -Xmx4G -jar $SNPEFFJAR GRCh37.75 -formatEff -classic \
   | bgzip -c > $NORMVCF
tabix -p vcf $NORMVCF

# load the pre-processed VCF into GEMINI
gemini load --cores 3 -t snpEff -v $NORMVCF $db

# query away
gemini query -q "select chrom, start, end, ref, alt, (gts).(*) from variants" \
             --gt-filter " == HET and \
                 == HET and \
                          gt_types.kid == HOM_ALT" \


If you use GEMINI in your research, please cite the following manuscript:

Paila U, Chapman BA, Kirchner R, Quinlan AR (2013)
GEMINI: Integrative Exploration of Genetic Variation and Genome Annotations.
PLoS Comput Biol 9(7): e1003153. doi:10.1371/journal.pcbi.1003153

Table of contents

comments powered by Disqus