Background

Flow cytometry is used to detect fluorescent signals for many different applications in the biological and biomedical sciences. For example, it is used clinically for the immunophenotyping and diagnosis of haematological diseases such as blood cancers as well as for more fundamental applications including cell cycle analysis, DNA content assays, cell sorting and high-throughput screening. However, current strategies for the analysis of cytometry data are limited by two major factors.

First, the fluorescent signal must be sufficiently bright compared to background to reliably and quantitatively detect the presence of the fluorophore. Indeed, the sensitivity of fluorescence detection has been noted as a key issue in the application of flow cytometry to clinical diagnosis and analysis. In addition, sensitive detection of weak fluorophores is vital to multiple applications in basic research including detection of endogenously encoded genetic reporters, study of subcellular compartments or extracellular objects such as exosomes and in applications requiring fluorescence resonance energy transfer (FRET). More sensitive detection of fluorescence will allow such applications to be performed with ease as well as a wide range of other experiments that are not currently possible.

Second, while many different fluorophores exist, their emission spectra are often broad and the number of different fluorophores that can be used simultaneously is limited by spectral overlap. Critically, this limits the use of cytometry in fields such as high-content screening where multiple phenotypes are measured simultaneously.

Technology Overview

The University of Exeter have developed a new approach using AI for the analysis of cytometry data that provides two major advantages over existing methods:

Increased sensitivity to weak fluorescent signals
Improved differentiation of spectrally overlapping fluorescent signals

These algorithms analyse data from all spectral filters present in the flow cytometer and apply an AI approach to effectively identify fluorescent events, even when the signal intensity is well into the range of background signals (). This method is transferable to any fluorophore and is applicable to any existing flow cytometer without the need for hardware modification. In addition, the algorithms can accurately discriminate between fluorophores with substantially overlapping emission spectra. For example, the most basic models can differentiate between mCherry and mKR signals with an accuracy of at least 97% (). This is in contrast to existing gating methods, which struggle to differentiate between similar emission spectra, even when utilising compensation correction. Furthermore, the approach is scalable to cell populations of complex mixtures of fluorophores.

In summary, this new algorithms will allow for the application of flow cytometry with increased sensitivity, enabling applications that are not currently possible using existing gating strategies. In addition, it will be possible to simultaneously use a much wider range of fluorescent reporters while maintaining the ability to effectively distinguish between them and may avoid the need for methods like compensation correction entirely. Finally, the algorithms can be readily adapted to a wide range of additional applications, making this a versatile and powerful tool for the advancement of flow cytometry.

Benefits

Increased sensitivity to weak fluorescent signals
Improved differentiation of spectrally overlapping fluorescent signals

Opportunity

The University of Exeter is now seeking a commercial partner to facilitate translation of this research into commercial application through either (1) co-developing this technology further or (2) licensing the intellectual property.

Next-Generation Software for Cytometry Data Analysis

Seeking