EASY Lab.

SVNN: an efficient pipeline for structural variations calling using neural networks

SVNN is a fast, and accurate structural variation calling pipeline for long-reads that takes raw reads as the input and detects structural variants of size larger than 50 bp. SVNN utilizes NGMLR, and Minimap2 as long read aligners, and Sniffles, and SVIM as SV callers. SVNN achieves up to 16% sensitivity improvement in comparison with state-of-the-art methods and is two times faster than a naive combination of state-of-the-art tools.

Installation:

SVNN requires the executables of Minimap2, NGMLR, Sniffles, SVIM, and samtools to be located in the bin folder. Precompiled binaries of these tools for Ubuntu x86_64 is available in the bin folder. In case these binaries do not work for your system, you can download their source code from GitHub and compile them. Their download links are as below:

NGMLR: https://github.com/philres/ngmlr

Minimap2: https://github.com/lh3/minimap2

SVIM: https://github.com/eldariont/svim

Sniffles: https://github.com/fritzsedlazeck/Sniffles

Samtools: http://www.htslib.org/download/

SVNN source codes are available here.

To install SVNN:

            cd SVNN
            make
            pip install -e .

Working with SVNN:

Usage: SVNN [option] -r <reference> -q <reads> -s1 <Sniffles_SR> -s2 <SVIM_SR>

            reference: Path to the reference genome

            reads: Path to the read files

            Sniffles_SR [int]: number of supporting reads for Sniffles

            SSVIM_SR [int]: number of supporting reads for SVIM

OPTIONS:

            -t [int]: Number of threads [4]

Example: svnn -r chr21_22.fa -q reads.fastq -s1 4 -s2 2