Analysis of flow cytometry data using an automatic processing tool

Abstract

In spite of recent advances in flow cytometry technology, most cytometry data is still analyzed manually which is labor-intensive for large datasets and prone to bias and inconsistency. We designed an automatic processing tool (APT) to rapidly and consistently define and describe cell populations across large datasets. Image processing, smoothing, and clustering algorithms were used to generate an expert system that automatically reproduces the functionality of commercial manual cytometry processing tools. The algorithms were developed using a dataset collected from CMV-infected infants and combined within a graphical user interface, to create the APT. The APT was used to identify regulatory T-cells in HIV-infected adults, based on expression of FOXP3. Results from the APT were compared directly with the manual analyses of five immunologists and showed close agreement, with a concordance correlation coefficient of 0.96 (95% CI 0.91-0.98). The APT was well accepted by users and able to process around 100 data files per hour. By applying consistent criteria to all data generated by a study, the APT can provide a level of objectivity that is difficult to match using conventional manual analysis. ' 2008 International Society for Advancement of Cytometry Key termsimmunophenotyping; clustering; population; T-cell SINCE the first experiments in identifying cell populations based on the fluorescence characteristics of individual cells were performed in the late 1960s (1), there has been a steady increase in both the scope and availability of this technology to the extent that the benchtop multiparametric flow cytometer is now a standard piece of equipment in an immunology laboratory. The essentials of immunophenotyping have been established for much of that time, as cells are stained with monoclonal antibodies conjugated to flurochromes with nonoverlapping emission peaks, and the intensity of fluorescence in each wavelength is detected (2). Recent years have seen a rapid expansion in the number of parameters that can be assessed on a single cell, and it is now possible to record the intensities of seventeen different flurochromes (3), although most laboratories are still using machines with considerably fewer detectors.While considerable advances have been made in the development of flurochromes and the means to detect them, developments in the tools for analysis of multiparametric data have not kept pace. After measuring fluorescence intensity, most systems output data to an attached computer where each parameter is recorded for each cell. Processing at this stage is based on a limited number of dedicated flow cytometry analysis packages. Nearly all analyses are dependent on the subjective assessment of the operator. An increasing number of studies involve the detection of many different populations and subpopulations in a large number of individuals who may be sampled over a considerable period of time. Under such circumstances, it is unlikely that an individual operator can consistently apply the same criter...

A Survey of Flow Cytometry Data Analysis Methods

“…For example, a quick look at the first row in Table 1 shows the design components used by Jeffries et al [45] in their analysis of FCM data. Moreover, if somebody is interested in designing or using automated gating approaches, he/she can quickly identify the studies that address automated gating of the FCM data by referring to the third column of Table 1.…”

Automatic B cell lymphoma detection using flow cytometry data

“…Many existing dimension-reduction approaches may cause useful information to be lost [8-13]. There have been several attempts to use machine learning technique to automate the gating process [14-20]. The most commonly used approach is the k-mean algorithm [21], which assigns a cell to its nearest cluster.…”

High-Throughput Secondary Screening at the Single-Cell Level

“…Bagwell et al 10 introduced biplot in 1985, but since one- and two-color data were essentially the only opportunities for multiparameter data, biplot was not adopted within the core of the field. A variety of developments have created a series of analytical tools such as Kolmogorov-Smirnov (KS) tests, 11 basic cluster analysis, 12 neural networks, 13 prediction limit methods, 14 logical data display (reviewed in Herzenberg et al 15 ), complex classification schemes, 16 utility of support vector machines (SVMs) in flow cytometry, 17 the concept of fluorescence barcoding for multiplexing cells, 18 distance measures, 19 clustering tools, 20 merging mixture moments, 21 and Gaussian mixture modeling, 22 among others.…”