Discovery and Genotyping of Novel Sequence Insertions in Many Sequenced Individuals
C++ Python Other
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
mrsfast @ 32dda6a

Pamir: Discovery and Genotyping of Novel Sequence Insertions in Many Sequenced Individuals

Pamir detects and genotypes novel sequence insertions in single or multiple datasets of paired-end WGS (Whole Genome Sequencing) Illumina reads by jointly analyzing one-end anchored (OEA) and orphan reads.

Table of contents

  1. Installation
  2. Commands Options
  3. Output Formats
  4. Example Commands
  5. Publications
  6. Contact & Support


Source code of Pamir can be downloaded from GitHub. To begin with, you will need to set up several external tools as described below.


Pamir relies on specific version of the following tools:

  1. g++ 4.9.0 or higher

  2. Python 2.7 or higher (for the package argparse)

  3. samtools 0.1.19 or higher

  4. Velvet 1.2.10 or higher

  5. BLAST 2.3.0+ or higher

    You also need to download the latest BLAST nt database to /your/path/to/ncbi-blast-2.5.0+/db/ (see Compilation and Configuration below) for contamination detection.

mkdir /dir/to/blast/db
cd /dir/to/blast/db
/path/to/blast/bin/ nt

Compilation and Configuration

To install Pamir, you need to first fetch Pamir from our git repository or download the corresponding compressed files.

git clone --recursive
cd pamir

Then you need to update pamir.config in pamir folder with your paths for the binaries samtools, velveth, velvetg, blastn and the blast database folder. Open pamir.config using your preferred text editor, and modify your paths for the binaries:


Make the project

pamir$ make -j

Now you are ready to go!


SSE4 for mrsFAST

If sse4 is not supported in your system, you need to disable the flag of mrsfast by modifying line 20 of Makefile to be

make with-sse4=no -C mrsfast

Commands Options

Pamir is designed to detect novel sequence insertions based on one-end anchors (OEA) and orphans from paired-end Whole Genome Sequencing (WGS) reads.

Note that reference genome is required for running Pamir in addition to sequencing or mapping data.

A typical command to start Pamir is

./ -p my_project -r ref.fa --files [filetype]=[filenames]

Project Name

To run Pamir you have to specify a project name such that Pamir will create a folder to store the results and intermediate files. You need to specify project name by -p.

Project Location

Pamir will create the folder under the current working directory unless absolute or relative paths are explicity stated otherwise.

For example, when the users run pamir in /my/current/working/location,

  1. -p my_project: the project folder will be /my/current/working/location/my_project
  2. -p ../../my_project: the project folder will be /my/current/my_project
  3. -p /another/drive: the project folder will be /another/drive

Data Preparation

Required Information

Two information are required for running Pamir:

  1. Reference Genome: You need to provide the reference genome in single fasta file by specifying the parameters -r or --reference.

  2. Masking File: You can provide a file for masking reference genome. For example, you can ask Pamir to ignore events in repeat regions by giving -m repeat.mask . When you only want to consider events in genic regions, use -m genic.region --invert-masker and Pamir will mask those regions not in the given file.

Read Length

Now Pamir only accepts WGS datasets in which two mates of all reads are of equal length.

For datasets with inconsistent read lengths, Pamir will automatically detect the minimum read length and trim all reads into the same length by specifying --crop in mrsfast.

Sequencing Data

Pamir can take either FASTQ and SAM files as its input. It has three different options to accept inputs:

  1. SAM/BAM by mrsFAST-best-search: A paired-end mapping result of your WGS data which satisfies the following conditions:

    1. Every mate has exact one primary mapping record.

    In general, best mapping results from most mapper should be valid input files for this option. Pamir does not assume the order of mapping records in SAM/BAM files, while it took Pamir less time and memory to process SAM/BAM files which are sorted by name (or all records from the same read are grouped together).

    For example, a best-mapping SAM file is a valid input file for Pamir through --files mrsfast-best-search=wgs.sam.

    You can also give multiple best-mapping files separated by comma, --files mrsfast-best-search=sample1.sam,sample2.sam,sample3.sam, or the path to the folder that includes the inputs files, --files mrsfast-best-search=/directory/to/sample_best_mapping_sam_files/.

  2. FASTQ: Pamir also accepts FASTQ format as the input data once it is a single gzipped file in which two (equal-length) mates of a read locate consecutively. You can specify by giving --files fastq=wgs.fastq.gz.

  3. Alignment file SAM/BAM: Pamir also accepts any other alignment output sorted by readname. Alignment output can be in SAM or BAM format. You can specify by --files alignment=wgs.sam or --files alignment=wgs.bam.

mrsFAST Parameters

Pamir uses mrsFAST for multi-mapping the orphan and OEA reads obtained from the best-mapping output. You can give your own mrsFAST parameters or Pamir will use the default values. Some of the parameters you may want to update are :

  1. --mrsfast-n: Maximum number of mapping loci of anchor of an OEA. Anchor with higher mapping location will be ignored. 0 for considering all mapping locations. (Default = 50)
  2. --mrsfast-threads: Number of the threads used by mrsFAST-Ultra for mapping. (Default = 4)
  3. --mrsfast-errors: Number of the errors allowed by mrsFAST-Ultra for mapping. In default mode Pamir does not give any error number to mrsFAST-Ultra, in which case it calculates the error value as 0.06 x readlength. (Default = -1; 6% of the read length)
  4. --mrsfast-index-ws: Window size used by mrsFAST-Ultra for indexing the reference genome. (Default = 12)

Other Parameters

  1. --num-worker: Number of independent prediction jobs to be created. You can define this parameter according to your core number. (Default = 1)
  2. --resume: Restart pipeline of an existing project from the stage that has not been completed yet.
  3. --assembler: The assembler to be used in orphan assembly stage (Options: velvet, minia, sga. Default = velvet).

Output Formats

Pamir generates a VCF file for detected novel sequence insertions. You can run genotyping for each sample after obtaining the VCF file by:

python projectFolder/insertions.out_wodups_filtered_setcov_PASS.sorted reference.fa.masked sample1_FASTQ_1.fq sample1_FASTQ_2.fq [readlength] [SAMPLENAME] [mrsfast-min] [mrsfast-max] [projectFolder] [TEMPLATE_LEN]

The default template length is set to be 1000.

Example Commands

  1. To start a new analysis from a mrsfast-best mapping result SAM file:
$ ./ -p my_project -r ref.fa --files mrsfast-best-search=sample.sam
  1. make a pooled-run with multiple samples separated by comma:
$ ./ -p my_project -r ref.fa  --files mrsfast-best-search=sample.sam,sample2.sam,sample3.sam
  1. To make a pooled-run with multiple samples which are in a folder called SAMPLEFOLDER:
$ ./ -p my_project -r ref.fa  --files mrsfast-best-search=SAMPLEFOLDER
  1. To start from another mapping tool's alignment result SAM/BAM file:
$ ./ -p my_project -r ref.fa --files alignment=sample.bam
  1. To start from a gzipped fastq file,
$ ./ -p my_project -r ref.fa --files fastq=sample.fastq.gz
  1. To ignore regions in a mask file (e.g., repeat regions),
$ ./ -p my_project -r ref.fa -m repeat.txt --files mrsfast-best-search=sample.sam
  1. To analyze events only in some regions of the reference genome (e.g., genic regions),
$ ./ -p my_project -r ref.fa -m genic.region --invert-mask  --files mrsfast-best-search=sample.sam
  1. To make sure that mrsFAST will not report the mapping locations of an OEA read more after the 30th location:
$ ./ -p my_project -r ref.fa --mrsfast-n 30 --files mrsfast-best-search=sample.sam
  1. To specify the core number for mrsFAST during multi-mapping of OEAs:
$ ./ -p my_project -r ref.fa  --mrsfast-threads 8 --files mrsfast-best-search=sample.sam
  1. To speed up the prediction process by defining the independent prediction jobs according to available core numbers:
$ ./ -p my_project -r ref.fa  --num-worker 20 --files mrsfast-best-search=sample.sam
  1. To specify the assembler as sga for orphan assembly and also the number of prediction jobs will be 20:
$ ./ -p my_project -r ref.fa  --num-worker 20 --assembler sga --files mrsfast-best-search=sample.sam
  1. To resume from the previously finished stage:
$ ./ -p my_project --resume

Some Invalid Commands

The following parameters are not accepted by Pamir:

  1. Project name is missing:
$./ -r ref.fa --files alignment=sample.sam
  1. Reference genome file is missing:
$./ -p my_project --files alignment=sample.sam
  1. Incorrect path of the mask file:
$./ -p my_project -m non-exist-mask-file --files alignment=sample.sam
  1. No input sequencing files:
$./ -p my_project -r ref.fa
  1. Multiple sequencing sources:
$./ -p my_project  -r ref.fa --files mrsfast-best-search=sample.sam fastq=sample2.fastq.gz


Discovery and genotyping of novel sequence insertions in many sequenced individuals. P. Kavak*, Y-Y. Lin*, I. Numanagić, H. Asghari, T. Güngör, C. Alkan‡, F. Hach‡. Bioinformatics (ISMB-ECCB 2017 issue), 33 (14): i161-i169, 2017.

Contact & Support

Feel free to drop any inquiry to pinarkavak at gmail dot com .