Pattern recognition in flow cytometry.

BACKGROUND: Analytical flow cytometry (AFC), by quantifying sometimes more than 10 optical parameters on cells at rates of approximately 10(3) cells/s, rapidly generates vast quantities of multidimensional data, which provides a considerable challenge for data analysis. We review the application of multivariate data analysis and pattern recognition techniques to flow cytometry. METHODS: Approaches were divided into two broad types depending on whether the aim was identification or clustering. Multivariate statistical approaches, supervised artificial neural networks (ANNs), problems of overlapping character distributions, unbounded data sets, missing parameters, scaling up, and estimating proportions of different types of cells comprised the first category. Classic clustering methods, fuzzy clustering, and unsupervised ANNs comprised the second category.We demonstrate the state of the art by using AFC data on marine phytoplankton populations. RESULTS AND CONCLUSIONS: Information held within the large quantities of data generated by AFC was tractable using ANNs, but for field studies the problem of obtaining suitable training data needs to be resolved, and coping with an almost infinite number of cell categories needs further research.     

Current and future applications of flow cytometry in aquatic microbiology

Flow cytometry has become a valuable tool in aquatic and environmental microbiology that combines direct and rapid assays to determine numbers, cell size distribution and additional biochemical and physiological characteristics of individual cells, revealing the heterogeneity present in a population or community. Flow cytometry exhibits three unique technical properties of high potential to study the microbiology of aquatic systems: (i) its tremendous velocity to obtain and process data; (ii) the sorting capacity of some cytometers, which allows the transfer of specific populations or even single cells to a determined location, thus allowing further physical, chemical, biological or molecular analysis; and (iii) high-speed multiparametric data acquisition and multivariate data analysis. Flow cytometry is now commonly used in aquatic microbiology, although the application of cell sorting to microbial ecology and quantification of heterotrophic nanoflagellates and viruses is still under development. The recent development of laser scanning cytometry also provides a new way to further analyse sorted cells or cells recovered on filter membranes or slides. The main infrastructure limitations of flow cytometry are: cost, need for skilled and well-trained operators, and adequate refrigeration systems for high-powered lasers and cell sorters. The selection and obtaining of the optimal fluorochromes, control microorganisms and validations for a specific application may sometimes be difficult to accomplish.     

Hyperspectral cytometry at the single-cell level using a 32-channel photodetector

Despite recent progress in cell-analysis technology, rapid classification of cells remains a very difficult task. Among the techniques available, flow cytometry (FCM) is considered especially powerful, because it is able to perform multiparametric analyses of single biological particles at a high flow rate-up to several thousand particles per second. Moreover, FCM is nondestructive, and flow cytometric analysis can be performed on live cells. The current limit for simultaneously detectable fluorescence signals in FCM is around 8-15 depending upon the instrument. Obtaining multiparametric measurements is a very complex task, and the necessity for fluorescence spectral overlap compensation creates a number of additional difficulties to solve. Further, to obtain well-separated single spectral bands a very complex set of optical filters is required. This study describes the key components and principles involved in building a next-generation flow cytometer based on a 32-channel PMT array detector, a phase-volume holographic grating, and a fast electronic board. The system is capable of full-spectral data collection and spectral analysis at the single-cell level. As demonstrated using fluorescent microspheres and lymphocytes labeled with a cocktail of antibodies (CD45/FITC, CD4/PE, CD8/ECD, and CD3/Cy5), the presented technology is able to simultaneously collect 32 narrow bands of fluorescence from single particles flowing across the laser beam in <5 μs. These 32 discrete values provide a proxy of the full fluorescence emission spectrum for each single particle (cell). Advanced statistical analysis has then been performed to separate the various clusters of lymphocytes. The average spectrum computed for each cluster has been used to characterize the corresponding combination of antibodies, and thus identify the various lymphocytes subsets. The powerful data-collection capabilities of this flow cytometer open up significant opportunities for advanced analytical approaches, including spectral unmixing and unsupervised or supervised classification.     

Leukemia Prediction Using Sparse Logistic Regression

We describe a supervised prediction method for diagnosis of acute myeloid leukemia (AML) from patient samples based on flow cytometry measurements. We use a data driven approach with machine learning methods to train a computational model that takes in flow cytometry measurements from a single patient and gives a confidence score of the patient being AML-positive. Our solution is based on an regularized logistic regression model that aggregates AML test statistics calculated from individual test tubes with different cell populations and fluorescent markers. The model construction is entirely data driven and no prior biological knowledge is used. The described solution scored a 100% classification accuracy in the DREAM6/FlowCAP2 Molecular Classification of Acute Myeloid Leukaemia Challenge against a golden standard consisting of 20 AML-positive and 160 healthy patients. Here we perform a more extensive validation of the prediction model performance and further improve and simplify our original method showing that statistically equal results can be obtained by using simple average marker intensities as features in the logistic regression model. In addition to the logistic regression based model, we also present other classification models and compare their performance quantitatively. The key benefit in our prediction method compared to other solutions with similar performance is that our model only uses a small fraction of the flow cytometry measurements making our solution highly economical.     

Comparison of different staining procedures for the flow cytometric analysis of U-937 cells infected with different Leishmania-species.

The human macrophage cell line U-937 infected with different Leishmania species, Leishmania mexicana amazonensis (Lma), Leishmania donovani (Ld) and Leishmania infantum (Li), was analyzed by flow cytometry (FCM). Leishmania spp. were labeled with different stains prior to the infection of the U-937 cells (BCECF-Am, PKH2-GL and SYTO 17) or after the infection (AO, FITC-conjugated monoclonal antibodies, PI). Infected cells were analyzed by flow cytometry, fluorescence microscopy and in parallel microscopically after Giemsa staining. The data obtained by these two methods were compared to decide which method is mostly appropriate for detection and estimation of the infection rate. Three fluorescent stains were suitable: BCECF-Am, SYTO 17 and FITC-conjugated MoAb with 0.02% digitonin. None of the vital stains gave evaluable results after 3 days of incubation.     

Flow cytometric analysis of estrogen receptor expression in isolated nuclei and cells from mammary cancer tissues.

BACKGROUND: Cellular expression of receptors for the hormones estrogen and progesterone in human mammary tumors is of diagnostic and prognostic value. Ligand binding assays have been replaced by immunohistochemical analysis of receptor expression. However, both of these techniques are slow, and in the ligand-binding assay it is difficult to measure heterogeneity of receptor expression in individual cells. Flow cytometry has been used extensively for monitoring the expression of cellular receptors in hematopoietic tumors but has been of limited value in the analysis of mammary tumors, which are difficult to disaggregate into single cells for flow analysis. Hormone receptors have a predominant nuclear localization, and it is relatively easy to isolate nuclei from paraffin-embedded archival tissues for flow cytometric analysis of receptor expression. METHODS: Thick sections from formalin-fixed paraffin-embedded archival mammary tumors were digested by different enzyme solutions for the isolation of single nuclei. Different fixatives were used to compare the results on subsequent staining of the nuclei for estrogen receptor (ER) expression. Double staining with propidium iodide and fluorescein isothiocyanate labeled secondary antibodies for ER expression was used for multiparametric analysis of ER and DNA content. RESULTS: Digestion of paraffin sections with low concentration of pepsin and detergents was ideal for isolation of single nuclei. Fixation with paraformaldehyde in the presence of Triton X-100 improved staining of the cells. Isolated nuclei had enhanced immunoreactivity compared with the whole cells, and subpopulations differing in reactivity could be identified in the nuclear fractions. Double staining of nuclei for ER expression and DNA content could allow for multiparametric analysis of these two important parameters. CONCLUSIONS: The procedures described can be used for processing of archival paraffin-embedded mammary tumors for monitoring of ER expression and aneuploidy. These two parameters have important diagnostic and prognostic significance in mammary tumors. Laser flow cytometry by providing multiparametric analysis can allow for correlation of these cellular markers with other important cellular and clinical parameters.     

WGCNA: an R package for weighted correlation network analysis

Background

Modelling the time-related behaviour of biological systems is essential for understanding their dynamic responses to perturbations. In metabolic profiling studies, the sampling rate and number of sampling points are often restricted due to experimental and biological constraints.

Results

A supervised multivariate modelling approach with the objective to model the time-related variation in the data for short and sparsely sampled time-series is described. A set of piecewise Orthogonal Projections to Latent Structures (OPLS) models are estimated, describing changes between successive time points. The individual OPLS models are linear, but the piecewise combination of several models accommodates modelling and prediction of changes which are non-linear with respect to the time course. We demonstrate the method on both simulated and metabolic profiling data, illustrating how time related changes are successfully modelled and predicted.

Conclusion

The proposed method is effective for modelling and prediction of short and multivariate time series data. A key advantage of the method is model transparency, allowing easy interpretation of time-related variation in the data. The method provides a competitive complement to commonly applied multivariate methods such as OPLS and Principal Component Analysis (PCA) for modelling and analysis of short time-series data.     

Integrating Genetic,Neuropsychological and Neuroimaging Data to Model Early-Onset Obsessive Compulsive Disorder Severity

We propose an integrative approach that combines structural magnetic resonance imaging data (MRI), diffusion tensor imaging data (DTI), neuropsychological data, and genetic data to predict early-onset obsessive compulsive disorder (OCD) severity. From a cohort of 87 patients, 56 with complete information were used in the present analysis. First, we performed a multivariate genetic association analysis of OCD severity with 266 genetic polymorphisms. This association analysis was used to select and prioritize the SNPs that would be included in the model. Second, we split the sample into a training set (N = 38) and a validation set (N = 18). Third, entropy-based measures of information gain were used for feature selection with the training subset. Fourth, the selected features were fed into two supervised methods of class prediction based on machine learning, using the leave-one-out procedure with the training set. Finally, the resulting model was validated with the validation set. Nine variables were used for the creation of the OCD severity predictor, including six genetic polymorphisms and three variables from the neuropsychological data. The developed model classified child and adolescent patients with OCD by disease severity with an accuracy of 0.90 in the testing set and 0.70 in the validation sample. Above its clinical applicability, the combination of particular neuropsychological, neuroimaging, and genetic characteristics could enhance our understanding of the neurobiological basis of the disorder.     

Optimized between-group classification: a new jackknife-based gene selection procedure for genome-wide expression data

Background  

A recent publication described a supervised classification method for microarray data: Between Group Analysis (BGA). This method which is based on performing multivariate ordination of groups proved to be very efficient for both classification of samples into pre-defined groups and disease class prediction of new unknown samples. Classification and prediction with BGA are classically performed using the whole set of genes and no variable selection is required. We hypothesize that an optimized selection of highly discriminating genes might improve the prediction power of BGA.     

Supervised principal component analysis for gene set enrichment of microarray data with continuous or survival outcomes

MOTIVATION: Gene set analysis allows formal testing of subtle but coordinated changes in a group of genes, such as those defined by Gene Ontology (GO) or KEGG Pathway databases. We propose a new method for gene set analysis that is based on principal component analysis (PCA) of genes expression values in the gene set. PCA is an effective method for reducing high dimensionality and capture variations in gene expression values. However, one limitation with PCA is that the latent variable identified by the first PC may be unrelated to outcome. RESULTS: In the proposed supervised PCA (SPCA) model for gene set analysis, the PCs are estimated from a selected subset of genes that are associated with outcome. As outcome information is used in the gene selection step, this method is supervised, thus called the Supervised PCA model. Because of the gene selection step, test statistic in SPCA model can no longer be approximated well using t-distribution. We propose a two-component mixture distribution based on Gumbel exteme value distributions to account for the gene selection step. We show the proposed method compares favorably to currently available gene set analysis methods using simulated and real microarray data. SOFTWARE: The R code for the analysis used in this article are available upon request, we are currently working on implementing the proposed method in an R package.     

Detection of endogenous and antibody-conjugated alkaline phosphatase with ELF-97 phosphate in multicolor flow cytometry applications

BACKGROUND: The fluorogenic alkaline phosphatase (AP) substrate 2-(5'-chloro-2'-phosphoryloxyphenyl)-6-chloro-4-(3H)-quinazolinone (ELF(R)-97 phosphate, for Enzyme-Labeled Fluorescence) has been used primarily in microscope-based imaging applications to detect endogenous AP activity, antigens and various ligands in cells and tissues, and nucleic acid hybridization. In a previous study, we demonstrated the applicability of ELF-97 phosphate for detecting endogenous AP activity by flow cytometry. In this study, we show that the spectral characteristics and high signal-to-noise ratio provided by the ELF-97 phosphate make it a useful label for immunodetection via flow cytometry. It can be combined with a variety of other fluorochromes for multiparametric flow cytometry analysis of both endogenous AP activity and intracellular and extracellular immunolabeling with AP-conjugated antibodies. METHODS: ELF-97 phosphate detection of endogenous AP activity in UMR-106 rat osteosarcoma cells was combined with intracellular antigen detection using Oregon Green 488 dye-conjugated secondary antibodies and DNA content analysis using propidium iodide (PI) or 7-aminoactinomycin D (7-AAD). ELF-97 phosphate detection of endogenous AP was also tested for spectral compatibility with a variety of other commonly used fluorochromes. ELF-97 phosphate was then used to directly label intracellular antigens via AP-conjugated antibodies, again combined with the analysis of DNA content using PI and 7-AAD. ELF-97 phosphate was also used to directly detect extracellular antigens. It was combined with Oregon Green 488 dye, phycoerythrin (PE), and PE-Cy5 dye-labeled antibodies for simultaneous four-color analysis. All samples were analyzed on a dual-beam flow cytometer, with UV excitation of the ELF-97 alcohol reaction product. RESULTS: Application of the ELF-97 phosphate to detect AP was found to be compatible with immunodetection and DNA staining techniques. It was also spectrally compatible with a variety of other fluorochromes. Endogenous AP activity could be detected simultaneously with both intracellular antigens labeled using Oregon Green 488 dye, PE, Cy5 dye and Alexa Fluor 568 dye-conjugated antibodies, and DNA content analysis with PI or 7-AAD. This multiparametric assay accurately delineated the distribution of AP in cycling cells and was able to identify cell subsets with varying endogenous AP levels. The ELF-97 alcohol reaction product was found to be an effective label for intracellular antigen immunolabeling with AP-conjugated reagents, and could also be combined with PI and 7-AAD. ELF-97 phosphate was also found to be a useful label for extracellular antigen immunolabeling with AP conjugates, and was compatible with Oregon Green 488 dye, PE, and PE-Cy5 dye-labeled antibodies for four-color surface labeling with minimal spectral overlap and color compensation. CONCLUSIONS: ELF-97 phosphate was shown to be a useful label for both endogenous and antibody-conjugated AP activity as detected by flow cytometry. Its spectral characteristics allow it to be combined with a variety of fluorochromes for multiparametric analysis. Cytometry 43:117-125, 2001. Published 2001 Wiley-Liss, Inc.     

Flow cytometric analysis of estrogen, progesterone receptor expression and DNA content in formalin-fixed, paraffin-embedded human breast tumors.

Flow cytometric analysis of estrogen (ER) and progesterone (PgR) receptor expression in archival human breast tumors is relatively difficult. We have used enzyme digestion and microwave antigen retrieval procedures for multiparametric flow cytometric analysis of ER and PgR expression and DNA content in nuclei isolated from formalin-fixed/paraffin-embedded primary breast tumors. Deparaffinized rehydrated tissue sections treated with pepsin were subjected to microwave irradiation for unmasking of ER and PgR antigenic sites. Biotinylated ER antibody and streptavidin-fluorescein isothiocyanate (FITC) were used for ER labeling and PgR antibody with phycoerythrin labeled goat anti-mouse antibody was used for PgR labeling. Counter staining with propidium iodide-RNase was used for determination of cellular DNA content. Our results show that enzyme digestion and microwave treatment of formalin-fixed, paraffin-embedded breast tumors can be successfully used for the multiparametric analysis of nuclear hormone receptor expression and DNA content by flow cytometry.     

OpenCyto: An Open Source Infrastructure for Scalable,Robust, Reproducible,and Automated,End-to-End Flow Cytometry Data Analysis

Flow cytometry is used increasingly in clinical research for cancer, immunology and vaccines. Technological advances in cytometry instrumentation are increasing the size and dimensionality of data sets, posing a challenge for traditional data management and analysis. Automated analysis methods, despite a general consensus of their importance to the future of the field, have been slow to gain widespread adoption. Here we present OpenCyto, a new BioConductor infrastructure and data analysis framework designed to lower the barrier of entry to automated flow data analysis algorithms by addressing key areas that we believe have held back wider adoption of automated approaches. OpenCyto supports end-to-end data analysis that is robust and reproducible while generating results that are easy to interpret. We have improved the existing, widely used core BioConductor flow cytometry infrastructure by allowing analysis to scale in a memory efficient manner to the large flow data sets that arise in clinical trials, and integrating domain-specific knowledge as part of the pipeline through the hierarchical relationships among cell populations. Pipelines are defined through a text-based csv file, limiting the need to write data-specific code, and are data agnostic to simplify repetitive analysis for core facilities. We demonstrate how to analyze two large cytometry data sets: an intracellular cytokine staining (ICS) data set from a published HIV vaccine trial focused on detecting rare, antigen-specific T-cell populations, where we identify a new subset of CD8 T-cells with a vaccine-regimen specific response that could not be identified through manual analysis, and a CyTOF T-cell phenotyping data set where a large staining panel and many cell populations are a challenge for traditional analysis. The substantial improvements to the core BioConductor flow cytometry packages give OpenCyto the potential for wide adoption. It can rapidly leverage new developments in computational cytometry and facilitate reproducible analysis in a unified environment.

This is a PLOS Computational Biology Software Article.

     

On multiparameter data analysis in flow cytometry

Increasing numbers of parameters that are accessible to simultaneous measurement in flow cytometric instruments, combined with the extremely large sample sizes common in flow cytometry, make it necessary to examine methods of multivariate statistics for their applicability to problems of visualization and quantitative analysis of flow cytometric data. This article describes some approaches to dimensionality reduction that appear well suited for data sets obtained by flow cytometry.     

Post-thaw evaluation of dog spermatozoa using new triple fluorescent staining and flow cytometry

A new triple fluorescent staining method was developed to evaluate frozen-thawed dog spermatozoa. This method was used to compare functional parameters of canine spermatozoa cryopreserved using 2 different freezing-thawing protocols. One ejaculate from each of 10 dogs was split into 2 aliquots and processed using the Andersen method or the CLONE method. Semen samples were evaluated immediately after thawing and after 3 h of incubation at 37 degrees C. Plasma membrane integrity and acrosomal status of the spermatozoa were evaluated simultaneously by flow cytometry using a combination of 3 fluorescent dyes: Carboxy-SNARF-1 (SNARF), to identify the live spermatozoa; propidium iodide (PI), which only stains dead cells or cells with damaged membranes; and fluorescein isothiocyanate (FITC)-conjugated Pisum sativum agglutinin (PSA), which binds to the acrosomal content of spermatozoa with damaged plasma and outer acrosomal membranes. The accuracy of this new staining method in quantifying the proportions of live and dead spermatozoa by flow cytometry was evaluated by comparing it with the staining technique using carboxyfluorescein diacetate and propidium iodide (CFDA-PI), which yielded high correlation coefficients. The triple-stained sperm samples were also analyzed by epifluorescence microscopy, and both methods proved to be highly correlated. Post-thaw progressive motility and plasma membrane integrity were similar for the 2 freezing procedures, but the proportion of damaged acrosomes after thawing was lower using the Andersen method and the spermatozoa had a higher thermoresistance. This new triple staining method for assessing canine sperm viability and acrosomal integrity provides an efficient procedure for evaluating frozen-thawed dog semen samples either by flow cytometry or fluorescence microscopy.     

Flow cytometry and cell sorting of heterogeneous microbial populations: the importance of single-cell analyses.

The most fundamental questions such as whether a cell is alive, in the sense of being able to divide or to form a colony, may sometimes be very hard to answer, since even axenic microbial cultures are extremely heterogeneous. Analyses that seek to correlate such things as viability, which is a property of an individual cell, with macroscopic measurements of culture variables such as ATP content, respiratory activity, and so on, must inevitably fail. It is therefore necessary to make physiological measurements on individual cells. Flow cytometry is such a technique, which allows one to analyze cells rapidly and individually and permits the quantitative analysis of microbial heterogeneity. It therefore offers many advantages over conventional measurements for both routine and more exploratory analyses of microbial properties. While the technique has been widely applied to the study of mammalian cells, is use in microbiology has until recently been much more limited, largely because of the smaller size of microbes and the consequently smaller optical signals obtainable from them. Since these technical barriers no longer hold, flow cytometry with appropriate stains has been used for the rapid discrimination and identification of microbial cells, for the rapid assessment of viability and of the heterogeneous distributions of a wealth of other more detailed physiological properties, for the analysis of antimicrobial drug-cell interactions, and for the isolation of high-yielding strains of biotechnological interest. Flow cytometric analyses provide an abundance of multivariate data, and special methods have been devised to exploit these. Ongoing advances mean that modern flow cytometers may now be used by nonspecialists to effect a renaissance in our understanding of microbial heterogeneity.     

Gene selection for sample classification based on gene expression data: study of sensitivity to choice of parameters of the GA/KNN method.

MOTIVATION: We recently introduced a multivariate approach that selects a subset of predictive genes jointly for sample classification based on expression data. We tested the algorithm on colon and leukemia data sets. As an extension to our earlier work, we systematically examine the sensitivity, reproducibility and stability of gene selection/sample classification to the choice of parameters of the algorithm. METHODS: Our approach combines a Genetic Algorithm (GA) and the k-Nearest Neighbor (KNN) method to identify genes that can jointly discriminate between different classes of samples (e.g. normal versus tumor). The GA/KNN method is a stochastic supervised pattern recognition method. The genes identified are subsequently used to classify independent test set samples. RESULTS: The GA/KNN method is capable of selecting a subset of predictive genes from a large noisy data set for sample classification. It is a multivariate approach that can capture the correlated structure in the data. We find that for a given data set gene selection is highly repeatable in independent runs using the GA/KNN method. In general, however, gene selection may be less robust than classification. AVAILABILITY: The method is available at http://dir.niehs.nih.gov/microarray/datamining CONTACT: LI3@niehs.nih.gov     

Data quality assessment of ungated flow cytometry data in high throughput experiments.

BACKGROUND: The recent development of semiautomated techniques for staining and analyzing flow cytometry samples has presented new challenges. Quality control and quality assessment are critical when developing new high throughput technologies and their associated information services. Our experience suggests that significant bottlenecks remain in the development of high throughput flow cytometry methods for data analysis and display. Especially, data quality control and quality assessment are crucial steps in processing and analyzing high throughput flow cytometry data. METHODS: We propose a variety of graphical exploratory data analytic tools for exploring ungated flow cytometry data. We have implemented a number of specialized functions and methods in the Bioconductor package rflowcyt. We demonstrate the use of these approaches by investigating two independent sets of high throughput flow cytometry data. RESULTS: We found that graphical representations can reveal substantial nonbiological differences in samples. Empirical Cumulative Distribution Function and summary scatterplots were especially useful in the rapid identification of problems not identified by manual review. CONCLUSIONS: Graphical exploratory data analytic tools are quick and useful means of assessing data quality. We propose that the described visualizations should be used as quality assessment tools and where possible, be used for quality control.