Benchmark dataset and analysis scripts for OrganelleCleaner: identification and removal of organellar contigs in genome assemblies

Version 2.0 of the benchmark dataset for the revised manuscript "OrganelleCleaner: identification and removal of organellar contigs in genome assemblies using graph topology, sequence features, and homology evidence" (Jeon and Kim). Results were produced with OrganelleCleaner v1.1.0 (https://github.com/Jeondong/OrganelleCleaner, tag v1.1.0; source snapshot included). The revised benchmark covers 10 plant species (Amborella trichopoda, Arabidopsis thaliana, Carica papaya, Citrullus lanatus, Glycine max, Malus domestica, Oryza sativa, Salix wilsonii, Solanum lycopersicum, Vitis vinifera; Daucus carota from version 1.0 was removed) and evaluates organellar contig detection against independent ground truth: Simulated assemblies with known answers: 14 hifiasm assemblies (HiFi and ONT) from reads simulated from T2T-level A. thaliana and O. sativa nuclear genomes plus organelle genomes at three organelle copy levels; each contig is labelled from the origin of its reads, and nuclear NUPT/NUMT regions are flagged as hard negatives. Distant organelle references: 53 target–reference species pairs from the same family to different seed-plant lineages. Merged-coverage proxy labels (supplementary): contigs whose merged BLAST coverage at ≥90% identity against NCBI plastid and mitochondrial genomes is ≥0.80. Contents: hifiasm contig and unitig graphs of the 10 real assemblies; simulated assemblies with truth tables and simulation references; own-species HiMT organelle references, proxy labels and distant references; OrganelleCleaner v1.1.0 reports for all real assemblies and modes; result tables behind the manuscript figures and tables; Vitis vinifera Hi-C scaffolding summaries; and all analysis scripts. See README.md for the file layout and methods. Raw sequencing data are publicly available at NCBI SRA, ENA and CNCB under the run accessions listed in README.md.The real-assembly benchmark files are not included because of their size; they can be regenerated from the public SRA/CNCB runs listed in Additional file 1 (Table S3) with hifiasm using default parameters.