AutoFlow: an interactive Shiny app for supervised and unsupervised flow cytometry analysis.

Area of Science:

• Immunology
• Computational Biology
• Bioinformatics

Background:

• Flow cytometry (FC) is crucial for cell analysis but manual thresholding is subjective and time-consuming.
• Advancements in FC technology necessitate automated analysis methods.
• Machine learning (ML) offers solutions but requires specialized expertise, highlighting the need for accessible tools.

Purpose of the Study:

• To develop an easy-to-use, open-source R Shiny application for automated flow cytometry analysis.
• To provide both supervised and unsupervised ML workflows for diverse FC data analysis needs.
• To enable scientists without extensive ML expertise to perform advanced FC analysis.

Main Methods:

• Developed AutoFlow, an R Shiny application integrating ML algorithms for FC data.
• Implemented automated preprocessing: fluorescence compensation, debris exclusion, single-cell identification, and marker gating.
• Included MFI quantification and downstream classification/clustering capabilities.
• Validated using public (Mosmann, Nilsson Rare) and novel (BM-MPS) datasets.

Main Results:

• AutoFlow demonstrated robust performance across multiple datasets.
• Supervised classification on BM-MPS achieved 97.2% accuracy.
• High sensitivity and specificity were achieved for rare populations (Mosmann Rare: 87.5% sensitivity; Nilsson Rare: 87.9% sensitivity).
• Unsupervised clustering identified biologically relevant cell populations, including novel candidates.

Conclusions:

• AutoFlow offers a fast, reproducible, and scalable solution for automated FC analysis.
• The application democratizes ML-based FC analysis for bench scientists.
• AutoFlow facilitates high-throughput studies and enhances the discovery of rare or unexpected cell types.