Correct misassemblies in genome assembly drafts using linked or long sequencing reads
Cut sequences at positions with few spanning molecules.
Written by Shaun Jackman, Lauren Coombe, Justin Chu, and Janet Li.
Shaun D. Jackman, Lauren Coombe, Justin Chu, Rene L. Warren, Benjamin P. Vandervalk, Sarah Yeo, Zhuyi Xue, Hamid Mohamadi, Joerg Bohlmann, Steven J.M. Jones and Inanc Birol (2018). Tigmint: correcting assembly errors using linked reads from large molecules. BMC Bioinformatics, 19(1). doi:10.1186/s12859-018-2425-6
Tigmint identifies and corrects misassemblies using linked (e.g. MGI’s stLFR, 10x Genomics Chromium) or long (e.g. Oxford Nanopore Technologies long reads) DNA sequencing reads. The reads are first aligned to the assembly, and the extents of the large DNA molecules are inferred from the alignments of the reads. The physical coverage of the large molecules is more consistent and less prone to coverage dropouts than that of the short read sequencing data. The sequences are cut at positions that have insufficient spanning molecules. Tigmint outputs a BED file of these cut points, and a FASTA file of the cut sequences.
Tigmint also allows the use of long reads from Oxford Nanopore Technologies. The long reads are segmented and assigned barcodes, and the following steps of the pipeline are the same as described above.
Each window of a specified fixed size is checked for a minimum number of spanning molecules. Sequences are cut at those positions where a window with sufficient coverage is followed by some number of windows with insufficient coverage is then followed again by a window with sufficient coverage.
Install Tigmint using Brew
brew install tigmint
Install Tigmint using Conda
conda install -c bioconda tigmint
Install Tigmint using PyPI
pip3 install tigmint
Run Tigmint using Docker
docker run -it bcgsc/tigmint
Install Tigmint from the source code
Download and extract the source code.
git clone https://github.com/bcgsc/tigmint && cd tigmint cd src make
curl -L https://github.com/bcgsc/tigmint/archive/master.tar.gz | tar xz && mv tigmint-master tigmint && cd tigmint cd src make
Install Python package dependencies
pip3 install intervaltree pybedtools pysam numpy
Install the dependencies of Tigmint
brew install bedtools bwa samtools brew tap brewsci/bio brew install minimap2
Install the dependencies of ARCS (optional)
brew tap brewsci/bio brew install arcs links-scaffolder
Install the dependencies for calculating assembly metrics (optional)
brew install abyss seqtk
To run Tigmint on the draft assembly
myassembly.fa with the reads
myreads.fq.gz, which have been run through
tigmint-make tigmint draft=myassembly reads=myreads
bwa mem -Cis used to copy the BX tag from the FASTQ header to the SAM tags.
samtools sort -tBXis used to sort first by barcode and then position.
To run both Tigmint and scaffold the corrected assembly with ARCS:
tigmint-make arcs draft=myassembly reads=myreads
To run Tigmint, ARCS, and calculate assembly metrics using the reference genome
tigmint-make metrics draft=myassembly reads=myreads ref=GRCh38 G=3088269832
To run Tigmint with long reads in fasta or fastq format (
myreads.fq.gz) on the draft assembly
myassembly.fa for an organism with a genome size of gsize:
tigmint-make tigmint-long draft=myassembly reads=myreads span=auto G=gsize dist=auto
minimap2 map-ontis used to align long reads from the Oxford Nanopore Technologies (ONT) platform, which is the default input for Tigmint. To use PacBio long reads specify the parameter
tigmint-makeis a Makefile script, and so any
makeoptions may also be used with
tigmint-make, such as
- The file extension of the assembly must be
.faand the reads
.fa.gzfor long reads), and the extension is not included in the parameters
reads. These specific file name requirements result from implementing the pipeline in GNU Make.
- The minimum spanning molecules parameter (
tigmint-cutis heavily dependent on the sequence coverage of the linked or long reads provided. When running Tigmint with long reads, use
Gto your assembly organism’s haploid genome size for this parameter to be calculated automatically, or explicitly set
spanto a specific number if you are interested in adjusting it. See Tips for more details.
tigmint-long, the maximum distance between reads threshold should be calculated automatically based on the read length distribution. This can be done by setting the parameter
tigmint: Run Tigmint, and produce a file named
tigmint-long: Run Tigmint using long reads, and produce a file named
arcs: Run Tigmint and ARCS, and produce a file name
metrics: Run, Tigmint, ARCS, and calculate assembly metrics using
abyss-samtobreak, and produce TSV files.
Parameters of Tigmint
draft: Name of the draft assembly,
reads: Name of the reads,
G: Haploid genome size of the draft assembly organism. Required to calculate
spanparameter automatically. Can be given as an integer or in scientific notation (e.g. ‘3e9’ for human) 
span=20: Number of spanning molecules threshold. Set
span=autoto automatically select span parameter (currently only recommended for
cut=500: Cut length for long reads (
longmap=ont: Long read platform;
ontfor Oxford Nanopore Technologies (ONT) long reads,
pbfor PacBio long reads (
window=1000: Window size (bp) for checking spanning molecules
minsize=2000: Minimum molecule size
as=0.65: Minimum AS/read length ratio
nm=5: Maximum number of mismatches
dist=50000: Maximum distance (bp) between reads to be considered the same molecule. Set
dist=autoto automatically calculate dist threshold based on read length distribution (
mapq=0: Mapping quality threshold
trim=0: Number of bases to trim off contigs following cuts
t=8: Number of threads
ac=3000: Minimum contig length (bp) for tallying attempted corrections. This is for logging purposes only, and will not affect the performance.
Parameters of ARCS
Parameters of LINKS
Parameters for calculating assembly metrics
ref: Reference genome,
ref.fa, for calculating assembly contiguity metrics
G: Size of the reference genome, for calculating NG50 and NGA50
- If your barcoded reads are in multiple FASTQ files, the initial alignments of the barcoded reads to the draft assembly can be done in parallel and merged prior to running Tigmint.
- When aligning linked reads with BWA-MEM, use the
-Coption to include the barcode in the BX tag of the alignments.
- Sort by BX tag using
samtools sort -tBX.
- Merge multiple BAM files using
samtools merge -tBX.
- When aligning long reads with Minimap2, use the
-yoption to include the barcode in the BX tag of the alignments.
- When using long reads, the minimum spanning molecule thresholds (
span) should be no greater than 1/4 of the sequence coverage. Setting the parameter
span=autoallows the appropriate parameter value to be selected automatically (this setting requires the parameter
- When using long reads, the edit distance threshold (
nm) is automatically set to the cut length (
cut) to compensate for the higher error rate and length. This parameter should be kept relatively high to include as many alignments as possible.
Using stLFR linked reads
To use stLFR linked reads with Tigmint, you will need to re-format the reads to have the barcode in a
BX:Z: tag in the read header.
For example, this format
@V100002302L1C001R017000000#0_0_0/1 0 1 TGTCTTCCTGGACAGCTGACATCCCTTTTGTTTTTCTGTTTGCTCAGATGCTGTCTCTTATACACATCTTAGGAAGACAAGCACTGACGACATGATCACC + FFFFFFFGFGFFGFDFGFFFFFFFFFFFGFFF@FFFFFFFFFFFF@FFFFFFFFFGGFFEFEFFFF?FFFFGFFFGFFFFFFFGFFEFGFGGFGFFFGFF
should be changed to:
@V100002302L1C001R017000000 BX:Z:0_0_0 TGTCTTCCTGGACAGCTGACATCCCTTTTGTTTTTCTGTTTGCTCAGATGCTGTCTCTTATACACATCTTAGGAAGACAAGCACTGACGACATGATCACC + FFFFFFFGFGFFGFDFGFFFFFFFFFFFGFFF@FFFFFFFFFFFF@FFFFFFFFFGGFFEFEFFFF?FFFFGFFFGFFFFFFFGFFEFGFGGFGFFFGFF
After first looking for existing issue at https://github.com/bcgsc/tigmint/issues, please report a new issue at https://github.com/bcgsc/tigmint/issues/new. Please report the names of your input files, the exact command line that you are using, and the entire output of Tigmint.