Characterization of the main strains of the ARC inoculant via machine learning-assisted single-cell Raman spectroscopy

Aspergillus-Rihizobia coupling (ARC) inoculant is a brand-new inoculant with coupling function that enhances legume quality and nitrogen fixation. Comprehensive characterization of its key functional strains is critical for establishing a quality-control framework for the inoculant’s formulation. Here, we constructed a characteristic spectral dataset comprising over 63,000 single-cell Raman spectra of the constituent strains by employing Ramanome technology. Six machine learning-based predictive models were developed and compared for the constituent strains, the Linear Discriminant Analysis (LDA) model demonstrated the best performance, with a classification accuracy exceeding 92.4%. This work provides a unique spectral fingerprint for the ARC inoculant and will directly aid its application in sustainable agricultural production.