Limits...
Multivariate analysis of flow cytometric data using decision trees.

Simon S, Guthke R, Kamradt T, Frey O - Front Microbiol (2012)

Bottom Line: For research on the host site, flow cytometry has become one of the major tools in immunology.After weighting the data according to their class probabilities, we created a total of 13,392 different decision trees for each given cytokine with different parameter settings.While some of the decision trees reflected previously known co-expression patterns, we found that the expression of some cytokines was not only dependent on the co-expression of others per se, but was also dependent on the intensity of expression.

View Article: PubMed Central - PubMed

Affiliation: Research Group Systems Biology/Bioinformatics, Leibniz Institute for Natural Product Research and Infection Biology - Hans Knöll Institute Jena, Germany.

ABSTRACT
Characterization of the response of the host immune system is important in understanding the bidirectional interactions between the host and microbial pathogens. For research on the host site, flow cytometry has become one of the major tools in immunology. Advances in technology and reagents allow now the simultaneous assessment of multiple markers on a single cell level generating multidimensional data sets that require multivariate statistical analysis. We explored the explanatory power of the supervised machine learning method called "induction of decision trees" in flow cytometric data. In order to examine whether the production of a certain cytokine is depended on other cytokines, datasets from intracellular staining for six cytokines with complex patterns of co-expression were analyzed by induction of decision trees. After weighting the data according to their class probabilities, we created a total of 13,392 different decision trees for each given cytokine with different parameter settings. For a more realistic estimation of the decision trees' quality, we used stratified fivefold cross validation and chose the "best" tree according to a combination of different quality criteria. While some of the decision trees reflected previously known co-expression patterns, we found that the expression of some cytokines was not only dependent on the co-expression of others per se, but was also dependent on the intensity of expression. Thus, for the first time we successfully used induction of decision trees for the analysis of high dimensional flow cytometric data and demonstrated the feasibility of this method to reveal structural patterns in such data sets.

No MeSH data available.


Related in: MedlinePlus

(A) IL-2 KO-day 21, (B) GM-CSF KO-day 21.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3316995&req=5

Figure 12: (A) IL-2 KO-day 21, (B) GM-CSF KO-day 21.

Mentions: We used a data set from intracellular cytokine staining of activated Th cells (Frey et al., 2011b). The cells were stained and analyzed for the expression of six cytokines as described in the following. DBA/1 mice in the age of 6–12 weeks were subcutaneously immunized at the base of the tail with recombinant glucose-6-phosphate isomerase (G6PI) in an emulsion containing also Freunds complete adjuvant as described (Bruns et al., 2009; Frey et al., 2010a,b, 2011a,b). At day 21 after immunization, the draining lymph nodes (inguinal, axillary, paraaortic) were aseptically removed and prepared to a single cell suspension. In addition, beside the wild type DBA/1 mice (WT) also interferon-gamma (IFN-γ) receptor knock-out DBA/1 mice (KO) were analyzed (Frey et al., 2011b) and we performed the analyses also for other time points (day 9 and day 21 after immunization). Altogether, we studied four conditions: WT-day 21 (standard condition) as well as the additional conditions WT-day 9, KO-day 21, and KO-day 9. The additional conditions have only been applied for the results shown in Figures 10–12 for a comparative study and to investigate the robustness of the results against experimental variations. For detection of antigen-specific cells by their CD154 expression (Kirchhoff et al., 2007), cells (1 × 107/ml in a 48 well plate) were restimulated with 20 μg/ml G6PI. Control samples were left unstimulated. The total restimulation time was 6 h and Brefeldin A (Sigma) at 5 μg/ml was added to all samples for the last 4 h to block cytokine secretion and to stabilize CD154 expression. These assay conditions have been determined to be optimal for a simultaneous detection of CD154 expression and cytokine production in antigen-specific CD4+ T helper cells. At the end of the restimulation period, cells were washed with ice-cold phosphate-buffered saline (PBS) and incubated with the fixable amine-reactive Aqua viability stain (Invitrogen) for 30 min on ice, fixed with 2% paraformaldehyde in PBS and permeabilized with 0.5% Saponin/0.5% BSA/0.02% NaN3 in PBS. Non-specific binding of antibodies was blocked by preincubation of the cells with anti-CD16/32 (2.4G2) and rat IgG (both at 5 μg/ml) for 8 min, followed by staining with fluorochrome-conjugated mAbs against CD4, CD154, GM-CSF, TNF-α, RANKL, IL-2, IL-17, and IFN-γ (all from BD, eBiosciences, Biolegend, or Miltenyi Biotech). For optimal staining results all antibodies were properly titrated and the binding of the anti-CD4 antibody to fixed and permeabilized cells was verified. After an additional washing step 0.5% Saponin/0.5% BSA/0.02% NaN3 in PBS, cells were resuspended in 0.5% BSA/0.02% NaN3 in PBS and measured within 3 h after staining. Cell analysis was performed on a BD LSR II flow cytometer equipped with 405, 488, and 633 nm laser lines and standard filter sets, except additional detectors for detection of Alexa-700 (red laser, 685 nm long-pass and 710/50 band-pass filters) and Qdot655 (violet laser, 635 nm long-pass and 670/14 band-pass filters, not used for this study). For fluorescence standardization and monitoring of the instrument performance, the cytometer setup, and tracking module of the BD FACSDiVa was used. Compensation for spectral overlap of the fluorochromes was done with the use of singly stained BD CompBeads and a compensation matrix was calculated using the BD FACSDiVa software. At least 1.5 million events were acquired.


Multivariate analysis of flow cytometric data using decision trees.

Simon S, Guthke R, Kamradt T, Frey O - Front Microbiol (2012)

(A) IL-2 KO-day 21, (B) GM-CSF KO-day 21.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3316995&req=5

Figure 12: (A) IL-2 KO-day 21, (B) GM-CSF KO-day 21.
Mentions: We used a data set from intracellular cytokine staining of activated Th cells (Frey et al., 2011b). The cells were stained and analyzed for the expression of six cytokines as described in the following. DBA/1 mice in the age of 6–12 weeks were subcutaneously immunized at the base of the tail with recombinant glucose-6-phosphate isomerase (G6PI) in an emulsion containing also Freunds complete adjuvant as described (Bruns et al., 2009; Frey et al., 2010a,b, 2011a,b). At day 21 after immunization, the draining lymph nodes (inguinal, axillary, paraaortic) were aseptically removed and prepared to a single cell suspension. In addition, beside the wild type DBA/1 mice (WT) also interferon-gamma (IFN-γ) receptor knock-out DBA/1 mice (KO) were analyzed (Frey et al., 2011b) and we performed the analyses also for other time points (day 9 and day 21 after immunization). Altogether, we studied four conditions: WT-day 21 (standard condition) as well as the additional conditions WT-day 9, KO-day 21, and KO-day 9. The additional conditions have only been applied for the results shown in Figures 10–12 for a comparative study and to investigate the robustness of the results against experimental variations. For detection of antigen-specific cells by their CD154 expression (Kirchhoff et al., 2007), cells (1 × 107/ml in a 48 well plate) were restimulated with 20 μg/ml G6PI. Control samples were left unstimulated. The total restimulation time was 6 h and Brefeldin A (Sigma) at 5 μg/ml was added to all samples for the last 4 h to block cytokine secretion and to stabilize CD154 expression. These assay conditions have been determined to be optimal for a simultaneous detection of CD154 expression and cytokine production in antigen-specific CD4+ T helper cells. At the end of the restimulation period, cells were washed with ice-cold phosphate-buffered saline (PBS) and incubated with the fixable amine-reactive Aqua viability stain (Invitrogen) for 30 min on ice, fixed with 2% paraformaldehyde in PBS and permeabilized with 0.5% Saponin/0.5% BSA/0.02% NaN3 in PBS. Non-specific binding of antibodies was blocked by preincubation of the cells with anti-CD16/32 (2.4G2) and rat IgG (both at 5 μg/ml) for 8 min, followed by staining with fluorochrome-conjugated mAbs against CD4, CD154, GM-CSF, TNF-α, RANKL, IL-2, IL-17, and IFN-γ (all from BD, eBiosciences, Biolegend, or Miltenyi Biotech). For optimal staining results all antibodies were properly titrated and the binding of the anti-CD4 antibody to fixed and permeabilized cells was verified. After an additional washing step 0.5% Saponin/0.5% BSA/0.02% NaN3 in PBS, cells were resuspended in 0.5% BSA/0.02% NaN3 in PBS and measured within 3 h after staining. Cell analysis was performed on a BD LSR II flow cytometer equipped with 405, 488, and 633 nm laser lines and standard filter sets, except additional detectors for detection of Alexa-700 (red laser, 685 nm long-pass and 710/50 band-pass filters) and Qdot655 (violet laser, 635 nm long-pass and 670/14 band-pass filters, not used for this study). For fluorescence standardization and monitoring of the instrument performance, the cytometer setup, and tracking module of the BD FACSDiVa was used. Compensation for spectral overlap of the fluorochromes was done with the use of singly stained BD CompBeads and a compensation matrix was calculated using the BD FACSDiVa software. At least 1.5 million events were acquired.

Bottom Line: For research on the host site, flow cytometry has become one of the major tools in immunology.After weighting the data according to their class probabilities, we created a total of 13,392 different decision trees for each given cytokine with different parameter settings.While some of the decision trees reflected previously known co-expression patterns, we found that the expression of some cytokines was not only dependent on the co-expression of others per se, but was also dependent on the intensity of expression.

View Article: PubMed Central - PubMed

Affiliation: Research Group Systems Biology/Bioinformatics, Leibniz Institute for Natural Product Research and Infection Biology - Hans Knöll Institute Jena, Germany.

ABSTRACT
Characterization of the response of the host immune system is important in understanding the bidirectional interactions between the host and microbial pathogens. For research on the host site, flow cytometry has become one of the major tools in immunology. Advances in technology and reagents allow now the simultaneous assessment of multiple markers on a single cell level generating multidimensional data sets that require multivariate statistical analysis. We explored the explanatory power of the supervised machine learning method called "induction of decision trees" in flow cytometric data. In order to examine whether the production of a certain cytokine is depended on other cytokines, datasets from intracellular staining for six cytokines with complex patterns of co-expression were analyzed by induction of decision trees. After weighting the data according to their class probabilities, we created a total of 13,392 different decision trees for each given cytokine with different parameter settings. For a more realistic estimation of the decision trees' quality, we used stratified fivefold cross validation and chose the "best" tree according to a combination of different quality criteria. While some of the decision trees reflected previously known co-expression patterns, we found that the expression of some cytokines was not only dependent on the co-expression of others per se, but was also dependent on the intensity of expression. Thus, for the first time we successfully used induction of decision trees for the analysis of high dimensional flow cytometric data and demonstrated the feasibility of this method to reveal structural patterns in such data sets.

No MeSH data available.


Related in: MedlinePlus