Data and analysis materials for: Deposited-assembly usability and raw-read rescue influence genomic cohort composition in a global Klebsiella pneumoniae collection

Abstract

This dataset supports the study “Deposited-assembly usability and raw-read rescue influence genomic cohort composition in a global Klebsiella pneumoniae collection.” It contains accession-level manifests, derived statistical tables, figure-source data, analysis code, software-version information, and documentation of genome-input provenance. The final analytical cohort comprises 42,589 genomes: 40,093 retained deposited assemblies and 2,496 raw-read-derived replacements. Both input groups were assessed using common post-assembly criteria. The archive includes BioProject mappings for all 42,589 genomes across 340 projects and descriptive project-level sensitivity analyses. Exact-category sensitivity endpoints differ from the grouped primary endpoints; equal-project-weighted analyses and project-bootstrap intervals are exploratory. Version 1.6.0 adds sequencing-platform annotations and recovered processing-command provenance corresponding to Supplementary Table S6. All 2,497 run accessions associated with the historical raw-derived input manifest were annotated by the European Nucleotide Archive as Illumina paired-end sequencing. Recovered fastp and Shovill commands establish input-to-output-path correspondence for 2,486 records. Complete fastp commands were not recovered for the remaining 11 records. These recovered commands document processing provenance and do not establish successful completion of every processing step or constitute a newly validated reconstruction pipeline. Of 44,278 eligible records, 1,670 were used in a separate study and excluded from this analysis; 17 were excluded because their deposited assemblies did not meet the input criteria and raw-read download attempts were unsuccessful. Record-level logs documenting these 17 exclusions were unavailable. The historical analyzable-input manifest contained 42,591 records; species adjudication excluded one genome from each input source, yielding the final 42,589-genome cohort. This version adds provenance documentation without changing cohort membership, primary statistical results, BioProject sensitivity results, or figure-source values. The study evaluates cohort composition and does not establish causal technical effects of reconstruction or assess paired deposited-versus-reconstructed genotype accuracy. Genome assemblies and raw reads were retrieved using scripts, download utilities, and manual downloads from repository webpages. The archive does not redistribute genome assemblies or raw sequencing reads. These data remain available from the original repositories through the accession identifiers supplied in the manifests. Jianhua Li and Jing Guo contributed equally. Jing Chen is the corresponding author (dongqingfei@126.com). Funding was provided by the Natural Science Foundation of Xiamen Municipality (3502Z202373092) and the Xiamen Medical and Health Guidance Project (3502Z20254ZD1079).