051 | Optimization of a combined spectral and imaging flow cytometry protocol integrated with AI for rapid human microbiota screening

Arianna Aquilini-Mummolo^1|2, Francesca D’Ascanio^1|2|3, Giulia Colasante^1|2, Domenico De Bellis^1|2, Pasquale Simeone⁴, Sabino Porro^1|2, Tamer Esmail^1|2, Ayesha Younas^1|2, Luana D’Onofrio^1|2, Isabel Vallejo Bermúdez^5|6, Camilla Cichella², Mariagiulia Filoso^1|2, Chiara Porro⁴, Paola Lanuti^1|2. | ¹Department of Medicine and Aging Sciences, University “G d’Annunzio”, Chieti-Pescara, Chieti; ²Center for Advanced Studies and Technology CAST, University “G d’Annunzio”, Chieti-Pescara; ³Department of Humanities, Law and Economics, “Leonardo da Vinci” University, Torrevecchia Teatina, Italy; ⁴Department of Clinical and Experimental Medicine, University of Foggia, Italy; ⁵Molecular Immunology Group CTS208, Department of Cell Biology, Physiology and Immunology, University of Cordoba, Cordoba, Spain; ⁶Immunology and Allergy Group GC01, Maimonides Biomedical Research Institute of Cordoba IMIBIC, University of Cordoba, Reina Sofia University Hospital, Cordoba, Spain.

The human microbiota is a complex ecosystem pivotal to immune regulation and metabolic health, yet current screening methods like metagenomics are often time-consuming and expensive. This study aimed to develop a rapid, high-throughput protocol combining spectral and imaging flow cytometry with artificial intelligence (AI) to characterize microbial populations at the single-cell level. Using Bifidobacterium, Lactobacillus, Helicobacter pylori, and Escherichia coli as model organisms, we optimized a staining panel selected from nine candidates, identifying DRAQ5 (DNA), CFSE (cytoplasm/proliferation), and Live/Dead Near-IR (viability) as the optimal combination based on signal-to-noise ratios. Samples were analyzed using the BD FACSymphony™ A5 SE and the BD FACSDiscover™ S8 Cell Sorter, platforms capable of resolving complex autofluorescence profiles and capturing morphological features. The resulting high-dimensional datasets were processed using AI algorithms, including LightGBM and XGBoost. The results demonstrated that this integrated approach could successfully distinguish between microbial species in complex mixtures with high specificity and sensitivity (LightGBM accuracy ~0.87), leveraging both spectral fingerprints and imaging data. This protocol overcomes the limitations of conventional cytometry and offers a cost-effective, scalable alternative to sequencing for detailed microbiota profiling in clinical and research settings.