.png){kind=icon}
.png){kind=icon}
Return
A project-aware and contamination-aware re-evaluation of host-unmapped reads in public Silene latifolia sequencing data
.png){kind=icon}
OA
AI
R
DOI：10.1186/s12864-026-13260-2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Reads that remain after host subtraction are usually discarded from host-centered plant sequencing projects, but they may contain residual host sequence, organellar fragments, microbial candidate signal, laboratory or reagent contamination, and unresolved genomic content. These components are especially difficult to interpret when public datasets combine reduced-representation and whole-genome sequencing designs. We reanalyzed 2,686 public Silene latifolia sequencing runs from five BioProjects. After fastp quality control and BWA-MEM2-based host subtraction, 801,976,057 reads remained unmapped, corresponding to 5.49% of reads before fastp filtering and 5.50% of reads retained after fastp filtering. The residual fraction varied substantially by BioProject and library design, from 0.54% to 6.80% of fastp-retained reads. Kraken2 classification using a plant-inclusive PlusPFP database revealed a threshold-sensitive Viridiplantae component. To validate this component beyond read-level labels, Viridiplantae-classified reads were recovered at four confidence thresholds, assembled with MEGAHIT when possible, and searched against nt with BLASTn. Depending on the confidence threshold, 1,870-2,074 runs yielded MEGAHIT contigs represented in the BLASTn output. Conservative top-hit classification showed that 45.3–50.2% of representative contigs were plant-like, 7.7–9.3% were non-plant-like, and 40.5–47.0% remained ambiguous or other. PRJNA907022 strongly dominated the BLASTn-represented contig set, contributing 98.22–99.91% of representative contigs. Genus-level microbial assignments included plant-associated candidates, pathogen-compatible signals, and human/laboratory-associated signals; therefore, major genera were reinterpreted using a contamination-aware framework inspired by Salter et al. rather than as confirmed plant residents. Public host-oriented plant sequencing data contain a measurable and structured residual read fraction rather than disposable noise. In S. latifolia, this fraction is best interpreted as plant-like-enriched in part, but taxonomically mixed, project-skewed, and contamination-sensitive. The study provides a reproducible framework for quantifying, validating, and cautiously interpreting residual reads from public plant sequencing data.
Keywords:
Silene latifolia
Unmapped reads
Host subtraction
Kraken2
MEGAHIT
BLASTn
Public sequencing data
Residual reads
Contamination-aware analysis
RAD-seq
Journal
IF:
3.7
Papers:
1.9W
Citations:
5.2W
Organization
Cited Papers