No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Discovery of Nuclear Mitochondrial Insertions (dinumt)


This software is designed to identify and genotype nuclear insertions of mitochondrial origin from whole genome sequence data. It consists of two programs: dinumt (di-nu-mite), which identifies sites of insertions in a single sample and gnomit (geno-mite), which genotypes those sites across multiple samples. There is an additional program named clusterNumtsVcf which will merge sites identified in multiple samples into a single merged file for genotyping.

Required third-party resources

A number of third party software packages are required by these programs:

In addition, you will need:

  • reference genome in fasta format (e.g. hs37d5.fasta)
  • individual MT sequence (e.g. MT.fa or chrM.fa)
  • bed file of annotated numts in reference (refNumts.bed for hg19 is included in package)

The genotyping step requires the use of a sample index file containing various sample-level information (mean insert size, coverage, etc). A template has been provided, and the relevant data can be obtained by using GATK (DepthOfCoverage walker) and Picard (CollectInsertSizeMetrics) or custom scripts.


Additional information about various parameters below:

  • --len_cluster_include : width of window to consider anchor reads as part of the same cluster, typically calculated as mean_insert_size + 3 * standard_deviation
  • --len_cluster_link : width of window to link two clusters of anchor reads in proper orientation, typically calculated as 2 * len_cluster_include
  • --max_read_cov : maximum read depth at potential breakpoint location, used to filter out noisy regions of the genome, typically calculated as 5 * mean_coverage
  • --output_support : output all sequence reads supporting an insertion event in SAM format to filename in --support_filename option
  • --mask_filename : bed file of all numts annotated in reference sequence, one is provided for GRCh37 but additional versions can be obtained from UCSC Genome Browser
  • --min_map_qual : mininum mapping quality required for anchor read to be considered for cluster support, default is 10 but can be adjusted as needed

Example workflow

An example workflow would be as follows: \
--mask_filename=refNumts.bed \
--input_filename=sample1.bam \
--reference=hs37d5.fa \
--min_reads_cluster=1 \
--include_mask \
--output_filename=sample1.vcf \
--prefix=sample1 \
--len_cluster_include=577 \
--len_cluster_link=1154 \
--insert_size=334.844984 \
--max_read_cov=29 \
--output_support \
grep ^# sample1.vcf > header.txt
cat *vcf | grep -v ^# | | --reference=hs37d5.fa > data.txt
cat header.txt data.txt > merged.vcf

(merged vcf can be split into smaller pieces with multiple sets of sites run in parallel, if need be) \
--input_filename=merged.vcf \
--mask_filename=refNumts.bed \
--info_filename=sampleInfo \
--output_filename=merged_geno.vcf \
--samtools=samtools \
--reference=hs37d5.fa \
--breakpoint \
--min_map_qual=13 \
--dir_tmp=/tmp \
--exonerate=exonerate \


Questions: Please contact Ryan Mills at 04/16/2014