A Supervised and Visualization-Enabled Toolkit for Robust DNA Reference Database Construction in eDNA Biomonitoring

Abstract Environmental DNA (eDNA) and DNA-based technologies are increasingly used for biodiversity assessment, but their taxonomic resolution depends strongly on the quality, completeness, and interoperability of reference databases. Public barcode repositories often contain heterogeneous sequence lengths, inconsistent taxonomy, redundant records, and incomplete metadata, and existing tools do not always provide supervised curation, visual inspection, controlled database extension, and multiformat export in one workflow. Here, we present MFRefData (Multi-Format Reference Database), a Python-based toolkit for constructing customized reference databases using a paired CSV–FASTA architecture. MFRefData integrates NCBI sequence retrieval, primer-specific in silico PCR, standardized taxonomic annotation, VSEARCH-based dereplication, rule-based filtering, controlled database extension, visualization, and export to multiple commonly used compatible formats. Using the Pearl River fishes under standardized input conditions, MFRefData retrieved more sequences and covered more species than the refdb and CRABS toolkits, while also achieving higher sequence retrieval efficiency and memory efficiency. In an in silico mock community of 40 fish species and three mitochondrial 12S primer sets, MFRefData produced more correct species-level assignments than databases generated by the refdb and CRABS toolkits. Applied to real eDNA samples, MFRefData recovered broader species-level biodiversity signals. Overall, MFRefData provides a supervised, visualization-enabled, and multiformat framework for improving reference database transparency, technical quality control, and reproducibility, while supporting more reliable species-level taxonomic assignments in DNA-based biodiversity assessment.