otter

A generic targeted local-assembler for regions of interest for (aligned) long-read data. It automaticaly adapts to the local sequencing error-rates and coverages, and hence can be used on data from both PacBio and ONT instruments. It employs a genotyper to enable comparison of the assembled sequences across multiple samples.

It was initially devoloped for genotyping tandem repeats across one or more genomes, but can also be used for other structural variants such as transposable elements.

If given whole-genome alignments, otter can be used to compare and genotype regions of interest across multiple de novo assemblies.

Required:

• A c++17 supporting compiler

Included during install:

• WFA2-lib
• hclust-cpp

Optional, required for downstream analysis (VCF-generation and joint-genotyping):

• samtools

git clone --recursive https://github.com/holstegelab/otter.git && cd otter
# build WFA2-lib dependency
cd include/WFA2-lib; make clean setup lib_wfa; cd ../../
mkdir build
make

See commands and version by running:

./build/otter

Given a BED-formatted file (regions.bed) and aligned long-reads (reads.bam), perform targeted local assembly for each region of interest, and will output all unique allele-sequences:

otter assemble -b regions.bed -R sample_name --fasta reads.bam > assembly.fa

If an indexed reference genome is provided (ref.fa), otter will perform local read re-alignment on-the-fly if it detects potential suboptimal alignments due to highly divergent sequence between the region of interest in the read and a reference:

otter assemble -b regions.bed -r ref.fa -R sample_name --fasta reads.bam > assembly.fa

You can view the corresponding sequences of individual reads per region by using the --reads-only parameter.

Although the above examples output in FASTA-format, it is recommended to output in SAM/BAM to enable VCF generation and joint-genotyping (see section below):

#note: samtools is required
otter assemble -b regions.bed -r ref.fa -R sample_name reads.bam | samtools view -bh | samtools sort > assembly.bam
samtools index assembly.bam
#generate VCF file
otter genotype -b regions.bed -r ref.fa assembly.bam > assembly.vcf

otter can perform joint-genotyping of target-regions across multiple samples (e.g. a cohort) to enable uniform representations of structural variants. You can do this in three steps:

1. Locally assemble regions of interest with otter in BAM-format (see above).

2. Generate a single merged-BAM:

samtools merge -pco cohort.bam sample1.otter.bam sample2.otter.bam sample3.otter.bam ...
samtools index cohort.bam

3. Perform joint-genotyping and output in VCF-format:

otter genotype -b regions.bed -r ref.fa merged.bam > merged.vcf

It is possible to use the same workflow above to genotype regions of interest in one or more aligned whole-genome de novo assemblies. The whole-genome alignments must be in BAM-format. To genotype:

#for FASTA-format
otter wgat -b regions.bed --fasta aligned_assembly.bam > aligned_assembly.fa
#for BAM-format
otter wgat -b regions.bed aligned_assembly.bam | samtools view -bh | samtools sort > aligned_assembly.bam

There is metadata available for each allele-sequence in the auxillarary tags (BAM or FASTA):

Additional tags are available for alignment-based output (e.g. wgat command or --reads-only parameter in assemble command):

Definitions:

• otter was developed for characterising structural variation, and hence it may not be able distinguish highly similar sequences (e.g. few nucleotide differences). In these cases, it will collapse the assembly into a single allele-sequence.
• The backbone of the assembly relies on spanning-reads, and hence there is a size limitation (spannable by the long-reads).

See our preprint (update coming soon)