The promise and challenges of characterising genome-wide structural variants: A case study in a critically endangered parrot

This is the GitHub Repository supporting our submission to the Molecular Ecology Resources Special Issue Omics Tools for Species Conservation.
DOI: https://doi.org/10.1111/1755-0998.13783

Here are the steps we took to trial different strategies for SV discovery using Illumina and Oxford Nanopore Technology (ONT) sequence data. The raw Illumina reads were preprocessed by Joseph Guhlin under the Kākāpō125+ Sequencing Consortium. Sequencing coverage for the Illumina data ranged from 9x - 38x with a mean of ~18x, and ranged from 4.2x - 15.8x with a mean of 9.7x coverage for the seven indivudals included for long-read based SV discovery.

For this manuscript, three short-read discovery tools (Delly, Manta and SMOOVE) and two long-read SV discovery tools (CuteSV and Sniffles) were used. We also implemented the short-read genotyping tool BayesTyper to genotype outputs from CuteSV, Manta and Sniffles. Below is a short summary of each of the steps taken in this study.

1) Illumina short-read alignment with BWA

All short read data were accessed through the kākāpō 125+ sequencing consortium. These processed reads were aligned to the kākāpō reference genome (VGP) using BWA. The W sex chromosome (present in females) was removed prior to alignments of males as the W is highly repetitive, and may contain regions that would result in misalignment of reads. Other programs used in this step include: SAMtools v1.16.

2) Illumina short-read alignment statistics

Programs used Mosdepth v0.3.3, qualimap v2.2.2 and SAMtools v1.16. Summaries of these outputs were visualised using MultiQC v1.13.

3) ONT basecalling and read QC

Reads were basecalled using Guppy v6.3.7 using dna_r9.4.1_450bps_sup basecalling model. Adapters were trimmed from raw sequence reads using Porechop v0.2.4, lambda DNA removed using NanoLyse v1.2.0 and filtered for a minimium Q-score of 10 and length of 5kb using NanoFilt v2.8.0. Raw and filtered read quality were assessed using NanoPlot v1.39.0.

4) ONT alignment and statistics

Reads were mapped using Winnowmap v2.03, mapping quality was assessed using qualimap v2.2.2, Mosdepth v0.3.3 and summaries visualised using MultiQC v1.13.

5) Structural Variant discovery using the short-read discovery tool Delly v0.8.7

Called and genotyped SVs with Delly. BCFtoools was also used in this step.

6) Structural Variant discovery using the short-read discovery tool Smoove v0.2.8

Called and genotyped SVs with the Smoove pipeline. BCFtoools was also used in this step.

7) Structural Variant discovery using the short-read discovery tool Manta

Called SVs using two strategies 1) Joint and 2) Batched. BCFtoools was also used in this step.

8) Genotyping Manta outputs with BayesTyper v1.5

Genotyped SVs using BayesTyper for both Joint and Batched Manta outputs. Other programs used included BCFtoools and KMC

9) Structural variant discovery using the long-read discovery tools CuteSV v1.0.11 and Sniffles v2.0.7

After initial SV discovery with either CuteSV or Sniffles calls were refined using Jasminev1.1.5.

10) Genotyping CuteSV and Sniffles outputs with BayesTyper v1.5

SVs called using the long-read data were genotyped using BayesTyper.

11) Identified consensus SV calls

Identified overlapping SVs with SURVIVOR v1.0.7. BCFtoools was also used in this step.

12) Summarising SV characteristics

Here is an R markdown outlining how SV size distributions, SV type frequency, relative frequency per individual taking into account generation and lineage were visualised. This file also outlines how DAPCs for each of the 6 methods were constructed.