A structural variant filtering and prioritization tool for long-read sequencing data

INTRODUCTION

🚧 _needLR3.2 is currently a beta version and actively under construction 🚧

Please contact jgust1@uw.edu with issues or suggestions.

needLR is a command line tool that uses Jasmine merging to compare a query structural variant (SV) vcf to our collection of 1000 Genomes Project (1KGP) samples sequenced by Oxford Nanopore Technologies long-read sequencing (ONT LRS). The output is a .txt file with detailed annotations about the genomic context, OMIM phenotype association, and ancestry-specific allele frequencies of each of the SVs in the query vcf.

There are 3 key concepts that drive this project:

• More than half of suspected Mendelian conditions remain molecularly unsolved after current clinical testing methods.
• Basepair-for-basepair, there is more genetic diversity associated with structural variants (SVs) than SNVs and indels combined.
• Traditional short read sequencing technologies are underpowered to resolve up to half of the SVs per genome.

Thus, we need a comprehensive catalog of long-read sequencing (LRS)-based SV calls from healthy individuals in order to filter for rare potentially disease-causing SVs in clinical cases.

The Miller Lab is actively using Oxford Nanopore Technologies (ONT) LRS to sequence samples from the 1000 Genomes Project (1KGP). Our recent preprint describes the first 100 samples in the cohort. This version of needLR incorporates SV calls made by Sniffles2 v2.2 for 150 1KGP samples.

Please cite our 2024 preprint:

^{Gustafson JA, Gibson SB, Damaraju N, et al. Nanopore sequencing of 1000 Genomes Project samples to build a comprehensive catalog of human genetic variation. Preprint. medRxiv. 2024;2024.03.05.24303792. Published 2024 Mar 7. doi:10.1101/2024.03.05.24303792}

WORKFLOW

needLR performs the following steps on an input query vcf:

1. Preprocess query sample vcf to match 1KGP vcf format (SVs >=50bp, FILTER=PASS, full chromosomes)
2. Run Jasmine merge on 150 1KGP samples and 1 query sample
3. Isolate merged SVs seen in query sample (common and unique)
4. Assign ancestry aware allele frequencies to each query SV based on 1KGP sample input
5. Annotate query sample SVs with genomic context, OMIM phenotype association, and Hardy-Weinberg equilibrium check

INSTALLATION AND SET UP

needLR requires an environment with the following dependencies:

jasmine v1.1.5 ¹
bedtools v2.31.1 ²
bcftools v1.19 ³

¹ _{Kirsche M, Prabhu G, Sherman R, et al. Jasmine and Iris: population-scale structural variant comparison and analysis. Nat Methods. 2023;20(3):408-417. doi:10.1038/s41592-022-01753-3}
² _{Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26(6):841-842. doi:10.1093/bioinformatics/btq033}
³ _{Danecek P, Bonfield JK, Liddle J, et al. Twelve years of SAMtools and BCFtools. Gigascience. 2021;10(2):giab008. doi:10.1093/gigascience/giab008}

Download the needLR_local directory from AWS into a parent directory you want to run needLR from

wget https://s3.amazonaws.com/1000g-ont/needLR/needLR_local.tar.gz

[!NOTE] The zip file is ~3.2G, the unzipped directory is ~9.5G

Extract the directory

tar -xvzf needLR_local.tar.gz

This will create a working directory called needLR_local. Everything needed to run needLR is inside.

Navigate into the working directory

cd needLR_local

This is what you should see:

backend_files/
needLR_3.2.sh 
needLR_output/
EXAMPLE/

RUN NEEDLR

The needLR_3.2.sh command takes 3 required arguments:

To run needLR:

needLR_3.2.sh -f /file/path/to/list.txt -g /file/path/to/reference/genome.fa -t 20

As needLR is running, you will see the Jasmine output building - it will look like this:

Then it will look like this:

And eventually it will look like this:

(Sometimes this step can take quite a while)

OUTPUT

Each query vcf will spawn an output subdirectory within the /needLR_local/needLR_output directory. The output files will be:

[!NOTE] needLR generates many temporary files when running, this can add up to ~2G

[!NOTE] Jasmine will automatically make a seperate /needLR_local/output directory (this can be ignored)

ANALYZING OUTPUT

The {SAMPLE_ID}_RESULTS.txt file can easily be opened in Excel. Be sure that Excel is set up to delimit columns by tab (and only tab)

These are the columns which can be sorted at will:

Column headers are also defined in /needLR_local/needLR_output/column_key.txt

EXAMPLE

In the downloaded zip file there is a directory EXAMPLE/ which contains an example vcf and an example input_list.txt.
If eveything is sut up correctly, you should be able to run the following command from inside the /needLR_local directory

./needLR_3.2.sh -f EXAMPLE/input_list.txt -g /backend_files/hg38_fasta/hg38.no_alt.fa -t 20

and find the output here: needLR_local/needLR_output/EXAMPLE_HG01511_sniffles2_needLR_3.2
The results should match the example output available here in Excel format.

Considerations/shortcomings:

All of the allele frequencies are currently based on the number of autosomes in the 1KGP sample set, rendering allele frequencies (other than 0) for SVs on chrX and chrY inaccurate.

Some SVs are called and pass all filters but have a genotype of “./.” This can lead to a SV with an allele frequency of 0, but a population frequency of >0.

PROOF OF CONCEPT

We ran needLR on sniffles2 vcfs from 8 individuals with pathogenic SVs that were unsolved by short-read sequencing.

needLR reduced the number of SVs per individual from an average of 22,583 total SVs to an average of 565 unique SVs:

Of the unique SVs, an average of 144 intersect with hg38 annotated genes:

Of the unique, genic SVs, even fewer intersect with canonical exons and/or are in genes associated with OMIM phenotypes:

ADITIONAL RESOURCES

Here is a "quick reference" annotated needLR output for just the first 150 1KGP samples (no query sample) - This has allele frequencies and genomic context info for each of the SVs seen in our first 150 samples.

COMING SOON

More 1KGP samples

Our goal is to have 500 1KGP samples sequenced and analyzed by the begining of 2025

needLR_Trio

For annotating inheritance and potential de novo SVs when trio data is available

chm13 compatibility

Additional phenotype annotations (ie.HPO terms)