Skip to content

Latest commit

 

History

History
48 lines (37 loc) · 2.17 KB

README.md

File metadata and controls

48 lines (37 loc) · 2.17 KB

Vulcan

Anaconda-Server Badge Anaconda-Server Badge

Installation:

conda install -c bioconda vulcan

program usage:

usage: vulcan [-h] -i INPUT [INPUT ...] -r REFERENCE -o OUTPUT [-w WORK_DIR] [-t THREADS] [-p PERCENTILE [PERCENTILE ...]] [-f] [-d] [-R] [-clr | -hifi | -ont | -any | -hclr | -hhifi | -hont | -cmd]


vulcan: map long reads and prosper🖖, a long read mapping pipeline that melds minimap2 and NGMLR

optional arguments:
-h, --help            show this help message and exit
-w WORK_DIR, --work_dir WORK_DIR
                        Directory of work, store temp files, default: ./vulcan_work
-t THREADS, --threads THREADS
                        threads, default: 1
-p PERCENTILE [PERCENTILE ...], --percentile PERCENTILE [PERCENTILE ...]
                        percentile of cut-off, default: 90
-f, --full            keep all temp file
-d, --dry             only generate config
-R, --raw_edit_distance
                        Use raw edit distance to do the cut-off
-clr, --pacbio_clr    Input reads is pacbio CLR reads
-hifi, --pacbio_hifi  Input reads is pacbio hifi reads
-ont, --nanopore      Input reads is Nanopore reads
-any, --anylongread   Don't know which kind of long read
-hclr, --humanclr     Human pacbio CLR read
-hhifi, --humanhifi   Human pacbio hifi reads
-hont, --humannanopore
                        Human Nanopore reads
-cmd, --custom_cmd    Use minimap2 and NGMLR with user's own parameter setting

Required arguments::
-i INPUT [INPUT ...], --input INPUT [INPUT ...]
                        input read path, can accept multiple files
-r REFERENCE, --reference REFERENCE
                        reference path
-o OUTPUT, --output OUTPUT
                        vulcan's output's prefix, the output will be prefix_{percentile}.bam

example:

./vulcan -r ./test/GCF_000146045.2_R64_genomic.fna -i ./test/test_reads.fa -w ./test/ -o vulcan