BioSynth Trust: Developing understanding and confidence in flow cytometry benchmarking synthetic datasets to improve clinical and cell therapy diagnos

点击翻译按钮获取中文摘要

Background Flow Cytometry (FC) is a ubiquitous biometrology tool for establishing the identity of cell populations within the clinical and biomanufacturing sectors. FC has three components that determine the metrological quality of the final answer - the upstream collection and preparation of the cell samples which may vary in integrity, the FC instrument itself and its inherent calibration/accuracy statement, and finally the subsequent data processing which typically involves some form of manual or automated cell cluster gating process.FC samples and analysis may range in dimensionality (the number of discrete unique cell types and clusters within a sample). Many practitioners would cite the FC instrument accuracy statements as evidence of metrology confidence. However, this ignores the subsequent data analysis and input of variation at this stage into the final cell count numbers.The analysis of the FC output data can be currently completed in two common manners - human manual gating and stratification of the clusters leading to final cell enumeration of the target cluster, and, by corollary, automated software solutions that purportedly remove the operator from most of the data analysis train thus reducing variation of output.The potential for variation in the data analysis is significant. Detailed work has already been completed that has specifically investigated the variation in human operators showing that as the complexity of the gating analysis increases, then the absolute variation of cell number output increases. A range of FC instrument manufacturers and third party automated unsupervised software solutions exist. Each software solution typically uses one of several different mathematical approaches for cluster identification, stratification and isolation. Recent research has specifically benchmarked different software solutions with the development and application of bespoke synthetic datasets. The benefit of synthetic datasets is that they allow for bespoke design of cell clusters; absolute cell number, distribution characteristics, skew, separation index, rare cell scenarios, etc, noting that the synthetic cell cluster design has been initially benchmarked against normal and rare live cell models.Project DetailThis project will be completed in collaboration with UK National External Quality Assessment Service for Leucocyte Immunophenotyping (UK NEQAS LI) - part of the Sheffield Teaching Hospitals NHS Foundation Trust. The aim of this new project is to develop enhanced confidence factors in the generation of synthetic cell cluster datasets and extend their performance capabilities and applicability. The project will consider and deliver to the following issues;- Development and benchmarking of synthetic datasets against specific disease states and live cell data.- Selection of disease states in collaboration with UK NEQAS LI - anticipating 2 to 3 different states to be used as exemplars.- Flow Cytometry based cell measurements at UK NEQAS LI to generate representative higher dimensionality cell cluster reference data (up to 12 FC fluorescence channels).- Design synthetic datasets with higher dimensionality.- Enhance confidence in the synthetic dataset generation process (including rare cell or Minimum Residual Disease) using manual and automated gating techniques.- Develop synthetic datasets that can be used for External Quality Assessment scheme (EQA) performance monitoring, matching and replacing live cell samples.- Monitor the use of synthetic cell cluster datasets within UK NEQAS LI participants during EQA testing rounds.- Explore the potential for using synthetic datasets for EQA participant training including rare disease states.- Develop a roadmap for synthetic dataset generation requirements to be used for the training of future potential artificial intelligence based FC data analysis systems