luozhhub / mutserve Goto Github PK

Mutserve is a variant caller for mtDNA to detect heteroplasmic sites in NGS data. It has been integrated in mtDNA-Server. For scalability reasons, mutserve is parallelized using Hadoop MapReduce but also available as a standalone tool.

• mutserve always reports the non-reference level as the heteroplasmy level, while mtDNA-Server reports the minor component.
• mutserve includes a Bayesian model for homoplasmy detection. It uses the 1000G Phase 3 data as a prior and calculates the most likely posterior probability for each genotype. mtDNA-Server only outputs homoplasmic variants with a coverage > 30.

You can run mutserve as a standalone tool starting with CRAM/BAM files and detecting heteroplasmic and homoplasmic sites. By default BAQ is set (--noBaq otherwise).

wget https://github.com/seppinho/mutserve/releases/download/v1.3.4/mutserve-1.3.4.jar

java -jar mutserve-1.3.4.jar  analyse-local --input <file/folder> --output <filename.vcf / filename.txt> --reference <fasta> --level 0.01

To create a VCF file as an output simple specify --output filename.vcf.gz. Please use this reference file when using BAQ.

Best Practice Pipelines recommend the following steps for BAM files preperation:

• Remove Duplicates (java -jar picard-tools-2.5.0/picard.jar MarkDuplicates),
• Local realignment around indels (GenomeAnalysisTK.jar -T RealignerTargetCreator, java -jar GenomeAnalysisTK.jar -T IndelRealigner)
• BQSR (GenomeAnalysisTK.jar -T BaseRecalibrator).

By default (--output filename does not end with .vcf or .vcf.gz) we export a TAB-delimited file including ID, Position, Reference, Variant & VariantLevel. Please note that the VariantLevel always reports the non-reference variant level. The output file also includes the most and second most base at a specific position (MajorBase + MajorLevel, MinorBase+MinorLevel). The reported variant can be the major or the minor component. The last column includes the type of the variant (1: Homoplasmy, 2: Heteroplasmy or Low-Level Variant, 3: Low-Level Deletion, 4: Deletion, 5: Insertion). See here for an example.

If you want a VCF file as an output, please specify --output filename.vcf.gz. Heteroplasmies are coded as 1/0 genotypes, the heteroplasmy level is included in the FORMAT using the AF attribute (allele frequency) of the first non-reference allele. Please note that indels are currently not included in the VCF. This VCF file can be used as an input for https://github.com/seppinho/haplogrep-cmd.

• The insertions/deletions calling is currently in beta.

As with v1.3.3 you can generate your gold standard given 2 variant files from the source files, which were used to generate mixtures in lab. With those files (call mutserve on the two samples) you can calculate the expected sites (parameter generate-gold) given a mixture ratio and subsequently compare the results from the lab-mixture to the this gold-standard with the performance parameter.

Provide the text files from mutserve output of the two files (the txt-variant files from analyse-local are used as input files - file1 for the major component and file2 for the minor mixture component), as well as the level of the mixture (value between 0 and 1) and the output file - which is the resulting gold-standard and input file for the next step - performance calculation - see below:

java -jar mutserve-1.3.4.jar  generate-gold --file1 <variantfileMajorComponent.txt> --file2 <variantfileMinorComponent.txt> --level <mixture levels (e.g. 0.01 for 1%)> --output <expectedvariants.txt>

If you have a mixture model generated, you can use mutserve for checking precision, specificity and sensitivity. The expected variants (homoplasmic and heteroplasmic) need to be provided as gold standard in form of a text file, with one column, containing the positions expected (this can now be calculated -see previous step). The txt-variant file from analyse-local is used as input file and length needs to be specified (typically 16569 for human mitochondrial genomes, but as there are different reference sequence, this can vary as well). The value provided in level indicates the threshold at which heteroplasmic levels are considered in the analysis.

java -jar mutserve-1.3.4.jar  performance --in <variantfile.txt> --gold <expectedvariants.txt> --length <size of reference (e.g. 16569)> --level <threshold for heteroplasmic levels (e.g. 0.01)>

If you use this tool, please cite this paper.

• git clone https://github.com/seppinho/mutserve
• Import Maven project into your favourite IDE
• maven install