Abstract
BACKGROUND: The skin microbiome is a critical component of dermatological health, with its dysbiosis implicated in conditions ranging from atopic dermatitis to cancer. Shotgun metagenomics offers an unparalleled resolution for comprehensive taxonomic and functional profiling, yet its application in dermatology is hampered by the high proportion of host DNA and the lack of consensus on best-practice bioinformatic pipelines. While Illumina's proprietary DRAGEN platform is widely used, its closed-source nature and cost limitations necessitate the validation of robust, open-source alternatives to democratize access and enable customization. METHODS: This study evaluates the performance of Kraken-based open-source pipeline as a viable alternative to the DRAGEN platform as well as the effect of currently available alignment-based de-hosting methods-Bowtie2, BWA, and Rsubread-to remove human DNA, assuring the use of highly-curated human reference genome thus avoiding the limitations of potentially incomplete or contaminated k-mer-based databases. By using shotgun metagenomic data from 83 healthy individuals we systematically compared the impact of these de-hosting procedures prior to Kraken2/DRAGEN taxonomic classification and functional profiling using HUMAnN 3.0 to assess the influence of methodological choices on skin microbial community composition and metabolic pathway abundance interpretation. RESULTS: Our analysis revealed marked discrepancies arising from the choice of de-hosting tool and taxonomic classifier, leading to substantial variability in microbial and functional profiles that could compromise clinical interpretation. Among the pipelines tested, Bowtie2 de-hosting combined with Kraken2 taxonomic classification and HUMAN functional profiling efficiently recovered well-established sex- and age-related bacterial associations in healthy skin that were missed by all other methods, including DRAGEN. This superior performance, together with its customizable features, underscores the value of this workflow for robust and clinically relevant dermatological metagenomic studies. CONCLUSIONS: Our findings underscore the decisive impact of bioinformatic pipeline selection on skin microbiome analysis and offer actionable guidance for reproducible and clinically meaningful research. We present a customizable workflow that enhances reproducibility and transparency while improving the translational value of metagenomic data. This approach strengthens the reliability of microbiome studies and supports the development of precision diagnostics and personalized therapeutic strategies in dermatology.