Kled: An ultra-fast and sensitive structural variant detection tool for long-read sequencing data

Introduction

Kled is designed to call SVs nicely and quickly using long-read sequencing data. It takes mapped reads file (bam) as input and reports SVs to the stdout in the VCF file format. Kled can yield precise and comprehensive SV detection results within minutes and can run on any modern computer without needing of any field knowledge of the user to perform the SV detection.

Compiling

Kled uses cmake build tools to build the project.

Make sure you have the following dependencies and cmake tools (>=3.15), g++ (gxx) (>=15.1):

• zlib >=1.3
• bzip2 >=1.0
• liblzma-devel >=5.8
• libcurl >=8.0
• openssl >=3.5
• libdeflate >=1.24
• libboost-devel >=1.84
• gmp >=6.3

To build the project, run:

git clone https://github.com/CoREse/kled
cd kled
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=/your/path ..
cmake --build . -j 16
cmake --install .

And you will have the kled built and kled and HapKled installed,

Conda

Kled is now on bioconda! To get kled, simply:

conda install kled -c bioconda

And that's all! But before that, be sure you will install kled in the right environment, to create a dedicated environment for kled:

conda create -n kled kled -c bioconda

Usage

Kled need a reference file (fasta) and at least one bam (sam/bam/cram) file that stores the mapped reads to call SVs, and output a VCF file to the standard output.

kled -R Refernce.fa Sample.bam > SVs.vcf

The default parameters are tuned for ONT data, if your inputs are CLR or CCS data, consider add --CLR or --CCS option to get a better result:

kled -R Reference.fa --CCS CCS.bam > SVs.vcf
kled -R Reference.fa --CLR CLR.bam > SVs.vcf

For the description of all parameters:

kled --help

HapKled usage

HapKled is a script that helps you handling kled with haplotype-tagged input.

Before running HapKled you should first compile (see compiling the haplotype-aware kled) the haplotype-aware kled and put the path of it to the environment variable HapAwareKled.

export HapAwareKled=/path/to/hap-aware-kled

And you also need an installed Clair3 and Whatshap, and export the path of the Clair3 models to environment variable Clair3ModelPath.

export Clair3ModelPath=/path/to/bin/models

HapKled need a reference file (fasta) and at least one bam (sam/bam/cram) file that stores the mapped reads to call SVs, and output a VCF file to the standard output.

HapKled -R Refernce.fa Sample.bam > SVs.vcf

For the description of all parameters:

HapKled --help

Change log

• 1.2.10:
  • Code clean.
  • Commented out the DEBUG macro.
  • Change the VCF date format.
  • Change the way to calculate the run hash.
  • Add test for cmake.

Version rule

Version format: kled vX.Y.Z[.pN], or kled version X.Y.Z[.pN]. Where X is the major version number, Y is the algorithm version, Z is the minor version, and N is the patch number.

X is updated when there is a major improvement of kled. Y is updated when there are updates that will influence the SV calling result, for example, the F1 benchmark score. Z is updated when the output VCF is changed given the same inputs (besides the running date). And pN is added when there are source code changes that do not influence the output of the build, i.e., versions with the same X.Y.Z shall have identical outputs (except for the run date) given the same inputs and parameters. X, Y, Z, and N all start with 1, and are increased by 1 on each update.

This rule is applied since kled version 1.2.10.

Citation

This work is published on Briefings in Bioinformatics, doi:10.1093/bib/bbae049, please visit the site for citations.

The HapKled is published on Frontiers in Genetics, doi:10.3389/fgene.2024.1435087, please visit the site for citations.