{
  "abstract": "Background Multi-color immunophenotyping assays are widely used to profile and characterize various cell biology phenomena, including cancer, immunology, and biologic drugs such as antibody therapies, adoptive cell therapies, and stem cell treatments. Understanding complex biological systems requires extensive testing conditions to gain deeper insights. High throughput flow cytometry enables rapid analysis of numerous samples stained with a broad array of fluorescent antibodies, but it introduces significant complexity in data analysis due to large data volumes, high sample counts, and numerous cell events, each measured across multiple markers.Traditional manual gating strategies in cytometry analysis are subjective, reliant on expertise, time-consuming, and inadequate for unraveling complex high throughput data due to slow processing speeds, limited biological perspectives, and the absence of advanced machine learning and data visualization techniques. To address these challenges, we developed an optimized analysis pipeline specifically designed for high throughput flow cytometry-based immunophenotyping data, facilitating rapid fingerprint identification for better prediction and quick action.Methods This pipeline integrates advanced unsupervised machine learning algorithms with optimized settings, including partition clustering for cell type identification, Fast-t-SNE for a comprehensive biological overview, marker enrichment modeling for phenotype identification, and hierarchical clustering for enhanced visualization. An embedded feedback loop allows user input to refine cell type analysis. Multi-layered, table-based visualization of cell events across all testing conditions enables swift comparison for potential fingerprint identification. The pipeline is compatible with cloud platforms like Microsoft Azure, eliminating computational bottlenecks and offering substantial benefits in data storage, access, and sharing.Results As proof of concept, we applied this pipeline to analyze a high-color T cell phenotyping dataset. Our goal was to identify optimal conditions for T cell expansion in adoptive cell therapy. We processed a complete 96-well plate dataset, which included samples from three patients. Each patient had 16 conditions tested in duplicate wells, resulting in approximately 700,000 total events with eight markers measured per event. Remarkably, this analysis was completed in about nine minutes, without the need for subjective manual gating or extensive biological expertise. Within minutes post-computation, we identified optimal conditions for expanding T central memory cell, a critical cell type for cell therapy ( figures 1, 2). These results were verified and validated against traditional cytometry analysis, which typically requires hours to days to complete.Conclusions This cloud-based pipeline offers an efficient tool for high-throughput, high-event-count, multi-dimensional complex data analysis, uncovering biological insights and enabling users to make informed decisions.Abstract 36 Figure 1Plate-based immunophenotype fingerprintAbstract 36 Figure 2Compare and identify optimal conditions for T central memory cell expansion",
  "authors": [
    {
      "affiliations": [
        "Sartorius, Albuquerque, NM, USA"
      ],
      "name": "Zhaoping Liu"
    }
  ],
  "title": "36 Optimized analysis pipeline for high throughput flow cytometry: turning large complex data into quick action",
  "uid": "98198f0c-759e-5a2a-89b3-76c9e3501dc7"
}