A sensitive Mitochondrial variant detection pipeline from WGS data
Project description
mity
mity
is a bioinformatic analysis pipeline designed to call mitochondrial SNV and INDEL variants from Whole Genome Sequencing (WGS) data. mity
can:
- identify very low-heteroplasmy variants, even <1% heteroplasmy when there is sufficient read-depth (eg >1000x)
- filter out common artefacts that arise from high-depth sequencing
- easily integrate with existing nuclear DNA analysis pipelines (mity merge)
- provide an annotated report, designed for clinicians and researchers to interrogate
Usage
mity -h
More detailed usage can be found docs/commands.md
Dependencies
- python3 (tested on 3.7.4)
- freebayes >= 1.2.0
- bgzip + tabix
- gsort (https://github.com/brentp/gsort)
- pyvcf
- xlsxwriter
- pandas
Installation
Installation instructions via Docker, pip, or manually are available in INSTALL.md
Example Usage
This is an example of calling variants in the Ashkenazim Trio.
mity call
First run mity call
on three MT BAMs provided in docs/test-files.md. CRAM files are supported.
We recommend always using --normalise
, or mity report
won't work:
mity call \
--prefix ashkenazim \
--output-dir test_out \
--region MT:1-500 \
--normalise \
test_in/HG002.hs37d5.2x250.small.MT.RG.bam \
test_in/HG003.hs37d5.2x250.small.MT.RG.bam \
test_in/HG004.hs37d5.2x250.small.MT.RG.bam
This will create test_out/normalised/ashkenazim.mity.vcf.gz
(and tbi file).
or, if using Docker:
docker run -w "$PWD" -v "$PWD":"$PWD" drmjc/mity call \
--prefix ashkenazim \
--output-dir test_out \
--region MT:1-500 \
--normalise \
test_in/HG002.hs37d5.2x250.small.MT.RG.bam \
test_in/HG003.hs37d5.2x250.small.MT.RG.bam \
test_in/HG004.hs37d5.2x250.small.MT.RG.bam
mity report
We can create a mity report
on the normalised VCF:
mity report \
--prefix ashkenazim \
--min_vaf 0.01 \
--output-dir test_out \
test_out/ashkenazim.mity.vcf.gz
This will create: test_out/ashkenazim.annotated_variants.csv
and test_out/ashkenazim.annotated_variants.xlsx
.
mity normalise
High-depth sequencing and sensitive variant calling can create many variants with more than 2 alleles, and in some
cases, joins two nearby variants separated by shared REF
sequence into a multi-nucleotide polymorphism
as discussed in the manuscript. Here, variant normalisation relates to decomposing the multi-allelic variants and
where possible, splitting multi-nucleotide polymorphisms into their cognate smaller variants. At the time of writing,
all variant decomposition tools we used failed to propagate the metadata in a multi-allelic variant to the split
variants which caused problems when reporting the quality scores associated with each variant.
Technically you can run mity call
and mity normalise
separately, but since mity report
requires a normalised
vcf file, we recommend running mity call --normalise
.
mity merge
You can merge a nuclear vcf.gz file and a mity.vcf.gz file thereby replacing the MT calls from the nuclear VCF (
presumably from a caller like HaplotypeCaller which is not able to sensitively call mitochondrial variants) with
the calls from mity
.
mity merge \
--prefix ashkenazim \
--mity_vcf test_out/ashkenazim.mity.vcf.gz \
--nuclear_vcf todo-create-example-nuclear.vcf.gz
Recommendations for interpreting the report
Assuming that you are looking for a pathogenic variant underlying a patient with a rare genetic disorder potentially caused by a Mitochondrial mutation, then we recommend the following strategy:
- tier 1 or 2 variants included in the 'commercial_panels' column
- tier 1 or 2 variants that match the clinical presentation and the phenotype in 'disease_mitomap', preferably those that are annotated with Confirmed evidence in the 'status_mitomap' column
- exclude common variants: anything linked to 'phylotree_haplotype', high 'phylotree_haplotype', high 'MGRB_frequency', high 'GenBank_frequency_mitomap'.
- consider any remaining tier 1 or 2 variants that may have a predicted impact on tRNA
- consider any remaining variants with high numbers of 'variant_references_mitomap'
- if you have analysed multiple family members, consider variants who's level of 'variant_heteroplasmy' match the disease burden
- tier 3 variants have low numbers of supporting reads, and should be considered with caution. However we have observed numerous tier 3 variants, especially in WGS from blood, that match the pathogenic allele known to be at much higher heteroplasmy in the affected tissue (this phenomenon is well established in the literature). Thus, if there are any tier 3 variants identified that match the patient's clinical presentation, then we recommend considering these as candidate variants and validating using an orthogonal clinically validated assay, preferably on the disease affected tissue.
Reference genomes
Human
mity
natively supports the analysis of the revised Cambridge Reference Sequence (rCRS, RefSeq ID NC_012920.1). The
rCRS used in most human reference genomes from NCBI (GRCh37, hs37d5, GRCh38) and hg38 from UCSC, where it is either
named chrM
, or MT
. The main exception in common use is the hg19
reference genome from UCSC, which used a different
sequence (RefSeq NC_001807) which differs in length by 2bp, and sharing 99% sequence homology (16530/16572 identities)
and 4 gaps. For now, mity call
supports the hg19 reference, but mity report
will not annotate variants properly, so
you should not use this part of the pipeline. We strongly recommend that for mitochondrial analysis, to use a reference
genome that uses the rCRS sequence.
- the mitochondrial genome: since the release of the UCSC hg19 assembly, the Homo sapiens mitochondrion sequence (represented as "chrM" in the Genome Browser) has been replaced in GenBank with the record NC_012920, the revised Cambridge Reference Sequence (rCRS). We have not replaced the original sequence, NC_001807, as chrM in the hg19 Genome Browser. However, files in the subdirectory p13.plusMT include NC_012920 as "chrMT", in addition to the original "chrM".
Reference | contig name | RefSeq ID | length | rCRS |
---|---|---|---|---|
GRCh37 | chrM | NC_012920.1 | 16569 bp | rCRS |
hs37d5 | MT | NC_012920.1 | 16569 bp | rCRS |
hg19 (UCSC) | chrM | NC_001807.4 | 16571 bp | no |
GRCh38 | chrM | NC_012920.1 | 16569 bp | rCRS |
Mouse
mity
call
and normalise
support the analysis of the mouse genome (mity call --reference mm10 ...
). mity report
currently only supports variant annotation to the human rCRS sequence.
Commonly asked Questions
Base quality score recalibration (BQSR)
Most of the development of mity
was tested on BAM files that had undergone GATK's BQSR method, which improves the
base qualities of each read.
In our experience, this reduced the quality score of most bases by ~10 points, indicating that the base qualities
straight out of the sequencer are generally inflated. As the GATK best practices guide no longer recommends BQSR, it's
reasonable to ask whether mity
can be run on BAM files straight out of the aligner.
mity
has a custom QUAL score, which depends on the base qualities of only the reads that support the alternative
allele.
For tier 1 or 2 variants, there will be so many supporting reads, that any miscalibration of base quality scores will
have no material effect. Tier 3 variants with very few supporting reads may be impacted, where a variant with only 3 or
4 supporting reads may end up having a stronger mity QUAL score than after BQSR. The comment above regarding how you
should interpret and validate tier 3 variant still holds.
We would appreciate any feedback you may have on this.
CRAM support
CRAM support was added to mity call
in v0.4.0.
Acknowledgements
We would like to thank:
- The Kinghorn Centre for Clinical Genomics and collaborators, who helped with feedback for running
mity
. - The Genome in a Bottle consortium for providing the test data used here
- Eric Talevich who's CNVkit helped us structure
mity
as a package - Erik Garrison for developing
FreeBayes
and his early feedback in optimisingFreeBayes
for sensitive variant detection. - Brent Pederson for developing
gsort
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.