Machine learning integrates metabolomics and proteomics to identify key regulators of anthocyanin biosynthesis in edible rose petals

Edible rose petals represent a promising source of anthocyanins, natural pigments with health-promoting properties suitable for functional food applications. However, the molecular mechanisms regulating anthocyanin biosynthesis in roses remain incompletely understood. Here, we employed an integrated metabolomic and proteomic approach to investigate anthocyanin profiles and their regulatory networks in three edible rose varieties: Rosa alba (RA), Rosa damascena (RD), and Rosa centifolia (RC). We identified a total of 13 anthocyanins, with RC exhibiting the highest total anthocyanin content—6.9% higher than RD and fivefold greater than RA. Compositional analysis revealed variety-specific accumulation patterns: RD was rich in delphinidin and peonidin derivatives, while RA predominantly accumulated pelargonidin-3-O-rutinoside. Proteomic analysis identified 9,924 proteins, and weighted gene co-expression network analysis (WGCNA) highlighted the MEblue module as strongly correlated with anthocyanin accumulation. By integrating a KNN-based machine learning model, we identified ten key structural proteins (e.g., RhUFGT, F3H, DFR) and several transcription factors (e.g., NAC , bZIP_2 , C3H ) as central regulators. Our findings elucidate the molecular basis of anthocyanin biosynthesis in edible roses and provide potential targets for breeding strategies aimed at enhancing their value as functional food ingredients.