MTG-Link is a local assembly tool dedicated to linked-reads. It leverages barcode information from linked-reads to assemble specific loci.
The main feature of MTG-Link is that it takes advantage of the linked-read barcode information to get a subsample of reads of interest for the local assembly of each sequence. It also automatically tests different parameters values and performs a qualitative evaluation of the obtained assemblies.

MTG-Link can be used for various local assembly use cases, such as intra-scaffold and inter-scaffold gap-fillings, as well as the reconstruction of the alternative allele of large insertion variants. Notably, the sequence to be assembled can be totally unknown (even in a related species such as in targeted assembly). The locus of interest is only defined by two coordinates on a reference genome, indicating its left and right flanking sequences, and the sequence in-between to be assembled is referred to as the target sequence.

It takes as input a set of linked-reads, the target flanking sequences and coordinates in GFA format (with the flanking sequences identified as ”segment” elements (S lines and the targets identified as ”gap” elements (G lines)) and an indexed BAM file obtained after mapping the linked-reads onto the reference genome.
It outputs the set of assembled target sequences in Fasta format, as well as an assembly graph file in GFA format, complementing the input GFA file with the obtained target sequences.

Presently, it is directly compatible with the following linked-reads technologies, given that the barcodes are reported using the BX:Z tag:

• 10X Genomics
• Haplotagging
• stLFR
• TELL-Seq

MTG-Link is a Inria Genscale and INRAE / INRIA tool developed by Anne Guichard.

• Biopython
• Gfapy
• Mummer
• Pathos
• Pysam
• Regex
• LRez
• MindTheGap

For more information on the LRez and MindTheGap tools:

• LRez: https://github.com/morispi/LRez
• MindTheGap: https://github.com/GATB/MindTheGap

MTG-link is distributed as a Bioconda package, which can be installed with:

conda install -c bioconda mtglink

Clone the MTG-Link repository with:

git clone --recursive https://github.com/anne-gcd/MTG-Link.git

Please make sure you have installed all the external dependencies. Alternatively, you can install them via the "requirements.txt" file.
To install a list of packages into a specified conda environment, do the following:
conda create --name <env> --file requirements.txt

You can test the installation of MTG-Link with the script test.py located in the test/ directory, that runs MTG-Link on several small toy datasets.
Please run the following command to try out MTG-Link on these datasets:

cd test/
./test.py

To make sure MTG-Link is running properly, all tests should "Pass".
If at least one test "Fail", either MTG-Link is not installed properly or there is an unsolved issue in the code. In this case, please use the issue form of GitHub.

The inputs of MTG-Link are the following:

• Linked-read sequencing files (FASTQ, BAM and index files)
• Target sequence definition (GFA file)

In the simplest case, when the target sequences to be assembled can be defined each by two coordinates on a sequence from the reference genome, you can generate the GFA file from a BED file, with the following command:

./utils/bed2gfa.py -bed targetSequenceCoordinates.bed -fa referenceGenome.fasta -out targetSequenceCoordinates.gfa 

• targetSequenceCoordinates.bed: BED file containing the locus coordinates
• referenceGenome.fasta: FASTA file containing the chromosome or scaffold sequence (reference genome)
• targetSequenceCoordinates.gfa: Output GFA file

In other local assembly use cases, such as intra-scaffold and inter-scaffold gap-fillings or the reconstruction of the alternative allele of large insertion variants, the input GFA can be generated from other file types (BED, FASTA, VCF), using scripts in the utils/ directory.

Prior to running MTG-Link, the LRez barcode index of the linked-reads FASTQ file has to be built. This can be done with the following command:

LRez index fastq -f readsFile.fastq.gz -o barcodeIndex.bci -g

• readsFile.fastq.gz: Linked-reads file (must be gzipped). Warning: the barcode sequence must be in the header (BX:Z tag)
• barcodeIndex.bci: output file where the LRez barcode index will be stored

MTG-Link can be run with the following command:

mtglink.py DBG -gfa gfaFile.gfa -bam bamFile.bam -fastq readsFile.fastq.gz -index barcodeIndex.bci 

• gfaFile.gfa: GFA file containing the coordinates of the targets to fill
• bamFile.bam: BAM file of the linked-reads mapped on the reference genome. Warning: the associated .bai file must exist
• readsFile.fastq.gz: Linked-reads file (must be gzipped). Warning: the barcode sequence must be in the header (BX:Z tag)
• barcodeIndex.bci: LRez barcode index of the FASTQ file

    -out OUTDIR             Output directory [default: ./MTG-Link_results]
    -line LINE              Line of GFA file input from which to start analysis (if not provided, start analysis from first line of GFA file input) [optional]
    -bxuDir BXUDIR          Directory where the FASTQ files containing the subsample of reads are located (1 file per target) (format of FASTQ files: xxx.bxu.fastq) [to provide if the read subsampling step has already been done for this dataset]
    -t THREADS              Number of threads to use for the Read Subsampling step [default: 1]
    -flank FLANKSIZE        Flanking sequences' size (bp) [default: 10000]
    -occ MINBARCOCC         Minimum number of occurrences in target flanking regions for a barcode to be retained in the union set [default: 2]
    -ext EXTSIZE            Size of the extension of the target on both sides (bp); determine start/end of local assembly [default: 500]
    -l MAXLENGTH            Maximum assembly length (bp) (it could be a bit bigger than the length of the target to fill OR it could be a very high length to prevent for searching indefinitely [default: 10000]
    -m MINLENGTH            Minimum assembly length (bp), by default 2*(-ext) bp [default: 1000]
    -k KMERSIZE             K-mer size(s) used for local assembly [default: [51, 41, 31, 21]]
    -a ABUNDANCETHRESHOLD   Minimal abundance threshold for solid k-mers [default: [3, 2]]
    --force                 To force search on all '-k' values provided
    -max-nodes MAXNODES     Maximum number of nodes in contig graph [default: 1000]
    -nb-cores NBCORES       Number of cores to use for the Local Assembly step (DBG assembly) [default: 1]
    -max-memory MAXMEMORY   Maximum memory for graph building (in MBytes) [default: 0]
    -verbose VERBOSITY      Verbosity level for DBG assembly [default: 0]
    --multiple              To return the assembled sequences even if multiple solutions are found (by default, if MTG-Link returns multiple solutions, we consider 'No Assembly' as it is not possible to know which one is the correct one)

NB: When using the --force option, the --multiple option cannot be used, as otherwise it would filter unique solutions obtained with different -k values.

The main outputs of MTG-Link are the following:

• Output GFA file: [input_GFA_name]_mtglink.gfa
  • it is an assembly graph file in GFA format, that complements the input GFA file with the obtained assembled target sequences.
• Output FASTA file: [input_GFA_name].assembled_sequences.fasta
  • it is a sequence file in FASTA format, that contains the set of assembled target sequences.

NB: All the output files and directories (including the intermediate files) of MTG-Link are detailed in input-output_files.md.

MTG-Link can be used for various local assembly use cases, such as the reconstruction of loci of interest, intra-scaffold and inter-scaffold gap-fillings, as well as alternative allele reconstruction of large insertion variants.

We provide documentation and examples of these different use cases with tiny datasets and all the command lines to run. The main differences between these use cases lie in how to generate the input GFA file, and also how to convert the output of MTG-Link in other formats:

• a single locus of interest: starting from a BED file with locus coordinates.
• intra-scaffold gap-filling: starting from stretches of 'N's in a genome file.
• inter-scaffold gap-filling: starting from a set of scaffolds of interest (performing both scaffolding and gap-filling).
• assembly of alternative allele of insertion variants: starting from a VCF file, containing insertion sites.

Guichard, A., Legeai, F., Tagu, D. et al. MTG-Link: leveraging barcode information from linked-reads to assemble specific loci. BMC Bioinformatics 24, 284 (2023). https://doi.org/10.1186/s12859-023-05395-w

To contact a developer, request help, or for any feedback on MTG-Link, please use the issue form of github: https://github.com/anne-gcd/MTG-Link/issues/new

You can see all issues concerning MTG-Link here.

If you do not have any github account, you can also send an email to Anne Guichard ([email protected]).