README.md

ContextScore

Filtering step for the ContextSV long-read structural variant (SV) caller, utilizing a Random Forest model trained on SV validation features. Assign confidence scores to SV datasets based on coverage, genomic context, and other important alignment features, then filter low-confidence SVs to increase the precision of the final callset. Genomic context is determined from annotations using ANNOVAR and UCSC databases.

Installation

conda install -c wglab -c bioconda -c conda-forge contextscore

# Or using mamba (faster dependency resolution):
mamba install -c wglab contextscore

ANNOVAR setup

ANNOVAR is required for annotations and must be installed separately.

These are the required ANNOVAR components for ContextScore:

• --annovar: directory containing annotate_variation.pl and table_annovar.pl
• --annovar-db: ANNOVAR database directory

User Workflow

contextscore --input input.vcf --output scored.vcf --sample-coverage 30 --buildver {hg38,hg19} \
	--annovar /path/to/annovar --annovar-db /path/to/humandb

Sources for additional annotations (under data/ directory):

File	Source	Description	Link
cytobands_hg{19,38}.txt	UCSC Genome Browser	Cytoband annotations for human genome builds hg19 and hg38	UCSC hg19 / UCSC hg38
hg{19,38}_segmental_duplications.bed	UCSC Genome Browser	Segmental duplication annotations for human genome builds hg19 and hg38	UCSC hg19 / UCSC hg38
phastcons100way_hg{19,38}.bed	UCSC Genome Browser	PhastCons conservation scores for human genome builds hg19 and hg38	UCSC hg19 / UCSC hg38
simple_repeats_hg{19,38}.bed	UCSC Genome Browser	Simple repeat annotations for human genome builds hg19 and hg38	UCSC hg19 / UCSC hg38
fragile_sites_hg38.bed / fragile_sites_hg19_liftover.bed	HumCFS	Fragile site annotations for human genome builds hg38 and hg19 (liftover)	HumCFS