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.

Citation

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

Description

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.

Installation

Install Tigmint using Brew

Install Linuxbrew on Linux or Windows Subsystem for Linux (WSL), or install Homebrew on macOS, and then run the command

brew install tigmint

Install Tigmint using Conda

conda install -c bioconda 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/releases/download/v1.2.4/tigmint-1.2.4.tar.gz | tar xz && cd tigmint-1.2.4
cd src
make

Dependencies

Install Python package dependencies

pip3 install intervaltree pybedtools pysam numpy

Tigmint uses Bedtools, minimap2, BWA, zsh and Samtools. These dependencies may be installed using Homebrew on macOS or Linuxbrew on Linux.

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

Usage

To run Tigmint on the draft assembly myassembly.fa with the reads myreads.fq.gz, which have been run through longranger basic:

tigmint-make tigmint draft=myassembly reads=myreads

bwa mem -C is used to copy the BX tag from the FASTQ header to the SAM tags.
samtools sort -tBX is 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 GRCh38.fa:

tigmint-make metrics draft=myassembly reads=myreads ref=GRCh38 G=3088269832

To run Tigmint with long reads in fasta or fastq format (myreads.fa.gz or 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-ont is 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 longmap=pb. The former calls minimap2 -x map-ont while the latter calls minimap2 -x map-pb instead.

Note

tigmint-make is a Makefile script, and so any make options may also be used with tigmint-make, such as -n (--dry-run).
The file extension of the assembly must be .fa and the reads .fq.gz (or .fa.gz for long reads), and the extension should NOT be included in the parameters draft and reads on the command line (otherwise you will get an error). These specific file name requirements result from implementing the pipeline in GNU Make.
The minimum spanning molecules parameter (span) for tigmint-cut is heavily dependent on the sequence coverage of the linked or long reads provided. When running Tigmint with long reads, use span=auto and set G to your assembly organism's haploid genome size for this parameter to be calculated automatically, or explicitly set span to a specific number if you are interested in adjusting it. See Tips for more details.
For 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 dist=auto.
The long-to-linked-pe step of tigmint-long uses a maximum of 6 threads

tigmint-make commands

tigmint: Run Tigmint, and produce a file named $draft.tigmint.fa
tigmint-long: Run Tigmint using long reads, and produce a file named $draft.cut$cut.tigmint.fa
arcs: Run Tigmint and ARCS, and produce a file name $draft.tigmint.arcs.fa
metrics: Run, Tigmint, ARCS, and calculate assembly metrics using abyss-fac and abyss-samtobreak, and produce TSV files.

Parameters of Tigmint

draft: Name of the draft assembly, myassembly.fa
reads: Name of the reads, myreads.fq.gz
G: Haploid genome size of the draft assembly organism. Required to calculate span parameter automatically. Can be given as an integer or in scientific notation (e.g. '3e9' for human) [0]
span=20: Number of spanning molecules threshold. Set span=auto to automatically select span parameter (currently only recommended for tigmint-long)
cut=500: Cut length for long reads (tigmint-long only)
longmap=ont: Long read platform; ont for Oxford Nanopore Technologies (ONT) long reads, pb for PacBio long reads (tigmint-long only)
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=auto to automatically calculate dist threshold based on read length distribution (tigmint-long only)
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.
SORT_OPTS: specify any options to be used for sort

Parameters of ARCS

c=5
e=30000
r=0.05

Parameters of LINKS

a=0.1
l=10

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

Tips

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 -C option 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 -y option 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=auto allows the appropriate parameter value to be selected automatically (this setting requires the parameter G as well).
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.
Each Tigmint (and Tigmint-long) step can be run separately through specifying the target using tigmint-make. For example, the bwa index step for Tigmint with linked reads can be launched with tigmint-make tigmint-index
- Tigmint steps\targets (for linked reads): tigmint-index, tigmint-align, tigmint-molecule, tigmint-cut
- Tigmint-long steps\targets: tigmint-long-estimate, tigmint-long-to-linked, tigmint-long-cut
In the command line, SORT_OPTS can be specified to add options used for sort

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

Support

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.

guijie2015/tigmint