SVision is a deep learning-based structural variants caller that takes aligned reads or contigs as input. Especially, SVision implements a targeted multi-objects recognition framework, detecting and characterizing both simple and complex structural variants from three-channel similarity images.
SVision is free for non-commercial use by academic, government, and non-profit/not-for-profit institutions. A commercial version of the software is available and licensed through Xi’an Jiaotong University. For more information, please contact with Jiadong Lin (jiadong66@stu.xjtu.edu.cn) or Kai Ye (kaiye@xjtu.edu.cn).
Please cite our paper "Lin, J., Wang, S., Audano, P.A. et al. SVision: a deep learning approach to resolve complex structural variants. Nat Methods (2022)." PDF
- MacOS, Big Sur (V11.6)
- Ubuntu (V20.04, including Windows Subsystem for Linux)
- CentOS Linux (V7.6.1810)
## Get latest source code
git clone https://github.com/xjtu-omics/SVision.git
cd SVision
## Create conda environment and install SVision
conda env create -f environment.yml
python setup.py install
docker pull jiadongxjtu/svision:latest
Please check the wiki page for more usage and output file format.
We provided support scripts used in this study to filter SVision calls at SVisionUtil, please follow instructions to filter your own calls.
-o OUT_PATH Absolute path to output
-b BAM_PATH Absolute path to bam file
-m MODEL_PATH Absolute path to CNN predict model
-g GENOME Absolute path to your reference genome (.fai required in the directory)
-n SAMPLE Name of the BAM sample name
-g
path to the reference genome, the index file should under the same directory. Please include all chromosomes you want to detect in the reference file. SVision only call SVs from chromosomes specified in the reference.
-m
path to the pre-trained deep learning model. NOTE: Please use -m svision-cnn-model.ckpt while running your data.
NOTE: If your input contains the alignment of assemblies, please activate the contig mode with --contig
.
The demo data is ./supports/HG00733.svision.demo.bam, which is extracted from whole genome sequencing HiFi data of HG00733 used in this study. We provided HG00733 whole genome calls used in this study, which is available at SVisionUtil. The HiFi data of HG00733 is generated by HGSVC in a recent study published at Science.
-
Download reference genome GRCh38
-
Download pretrained CNN model. There are three files and please put all files under a directory (e.g., /home/user/svision_model/).
Before running, please create a directory for SVision output (e.g., /home/user/svision_out).
SVision -o ./home/user/svision_out -b ./supports/HG00733.svision.demo.bam -m /home/user/svision_model/svision-cnn-model.ckpt -g /path/to/reference.fa -n HG00733 -s 5 --graph --qname
docker run -v /local/path:/container/path jiadongxjtu/svision:latest SVision -o /container/path/svision_out -b /container/path/HG00733.svision.demo.bam -m /container/path/svision_model/svision-cnn-model.ckpt -g /container/path/reference.fa -n HG00733 -s 5 --graph --qname
REMINDER: For Docker run, please put your BAM file, reference file and the pre-trained model under the /local/path.