The field of flow cytometry continues to progress toward the goal of collecting and analyzing more data. Here we explore two recent advancements: spectral analyzer platforms to collect more data, and machine learning algorithms to analyze larger datasets.
Machine Learning Data Analysis
Machine learning is a branch of artificial intelligence in which systems can learn from large data sets, identify complex patterns and make intelligent decisions or educated predictions regarding future outcomes. As machine learning models become exposed to new data, they will adapt independently without the need for human interaction. Advances in machine learning algorithms and the increased availability of more powerful computers have led to the adoption of machine learning tools for large data set analysis in many fields, from automated financial trading to driverless smart cars. Researchers are also increasingly using machine learning to analyze multi-dimensional flow cytometry data. This allows for high throughput analysis beyond the fundamental measurement limits of cytometers and enables rapid screening and characterization to identify patterns across multiple markers. Machine learning has also been used to improve label-free single-cell analysis. One recent study used multistage machine learning to allow for label-free quantification, suggesting this may be useful when labeled biomarkers are not readily available, are cost-prohibitive or could disrupt cell behavior. An open-source workflow for analyzing imaging flow cytometry data using machine learning has even been made available by a team at the Broad Institute to “enable the scientific community to leverage the full analytical power of IFC-derived data sets” to reveal otherwise unappreciated populations of cells based on features that may be hidden to the human eye”. In clinical labs, machine learning will likely assist in repetitive or complex analysis, offering a potential solution for the current time consuming and error-prone process of manual interpretation.
Spectral analyzers are the next generation of multicolor flow cytometers, offering unparalleled sensitivity and detection range. Unlike conventional flow cytometers which use dichroic mirrors and bandpass filters to block, reflect, or transmit a photon based on its wavelength for detection on photomultiplier tubes, spectral analyzers incorporate different optical approaches to achieve higher resolution with the same amount of light being distributed over more detector elements. One approach uses gratings or prisms to disperse photons according to wavelength across a detector array. A more advanced approach passes light through an optical filter-based coarse wavelength division multiplexing (CWDM) demultiplexer arrays and is then collected on avalanche photodiodes. This design allows for an increased number of detectors per laser without sacrificing the sensitivity of each detector channel with narrow emission bands. This sensitivity allows spectral analyzers to perform applications that are not possible on conventional flow cytometers. For example, spectral analyzers are can be used to separate intrinsic cell autofluorescence from signals of extrinsic fluorescence probes. They are also useful for separating labels with narrow emission spectra, such as quantum dots or Raman scattering labels. There are now several commercially available spectral analyzers on the market from Sony Biosciences and Cytek. FluroFinder currently supports spectral machines and is working to add more features to complement the capabilities of spectral analyzers to advance multi-color flow cytometry. If you are interested in feature updates from FluoroFinder to stay on top of new products and capabilities to optimize your panel design, sign-up for updates here.
Flow cytometry remains one of the most popular laboratory techniques around. Advancements over the years have enabled researchers to study upwards of 20 parameters within a single sample, making modern-day flow cytometry experiments a far cry from their early predecessors. It is impossible to guess the level of complexity flow cytometry will reach, but it seems fair to say that the evolution of this powerful technique is far from over.