Integrated HG002 structural variant callsets for MLkhor false-positive pruning

This dataset contains two integrated structural variant (SV) callsets for the Genome in a Bottle HG002/NA24385 sample. These files support the development, testing, and benchmarking of MLkhor, a machine learning workflow for false-positive pruning of integrated SV call sets. The two VCF files provide integrated multi-technology SV callsets generated from long-read sequencing, short-read sequencing, and optical genome mapping data: HG002.ONT_SR_OGM_merged.vcf.gzIntegrated callset from Oxford Nanopore Technologies (ONT), Illumina short-read, and Bionano optical genome mapping (OGM) data. HG002.PB_SR_OGM_merged.vcf.gzIntegrated callset from PacBio/HiFi, Illumina short-read, and Bionano optical genome mapping (OGM) data. For both callsets, the long-read callers are DeBreak, Dysgu, NanoSV, SVIM-asm, and SVIM. The short-read callers are CNVnator and Manta. Bionano optical genome mapping calls are also included. The callsets were normalized, merged, and integrated using SVkhor, which performs SV normalization, within-modality merging, and cross-technology integration. These files serve as external data resources for MLkhor and are distributed separately from the MLkhor source code because of their size. They can be used to reproduce false-positive pruning experiments, generate candidate BED or TSV tables, benchmark pruning models, and test MLkhor workflows on integrated HG002 SV callsets. The callsets are research-use datasets derived from public reference-sample data. They are not intended for clinical diagnosis.