The accurate identification of rare antigen-specific cytokine positive cells from peripheral blood mononuclear cells (PBMC) after antigenic stimulation in an intracellular staining (ICS) circulation cytometry assay is challenging as cytokine positive events may be fairly diffusely distributed and lack an obvious separation from your negative population. suffers from subjectivity and Desmopressin inconsistency across different circulation operators. The use of statistical clustering methods does not remove the need to find an objective threshold between between positive and negative events since consistent identification of rare event subsets is usually highly challenging for automated algorithms especially when there is distributional overlap between the positive and negative events (“smear”). We present a new approach based on the measure that is much like manual thresholding in providing a hard cutoff but has the advantage of being decided objectively. The overall performance of this algorithm is compared with results obtained by expert visual gating. Several Rabbit polyclonal to NR3C1. ICS data units from your External Quality Assurance Program Oversight Laboratory (EQAPOL) proficiency program were used to make the comparisons. We first show that visually decided thresholds are hard to reproduce and present a problem when comparing results across operators or laboratories as well as problems that occur with the use of commonly employed clustering algorithms. In contrast a single parameterization for the method performs consistently across different centers samples and instruments because it optimizes the precision/recall tradeoff by Desmopressin using both negative and positive controls. thresholds to expert visual gating optimized using back-gating by making use of data from your multi-center proficiency study EQAPOL 2 Methods 2.1 Data units Two 11-color data units (11C-EQAPOL-1 11 with explicit positive (SEB) stimulations were used in this study as well as a 4-color data set (4C-EQAPOL) without an explicit positively stimulated control. Negative controls for the 11-color data included co-stimulatory monoclonal antibodies (mAbs) anti-CD28 and anti-CD49d together with both Brefeldin A (BRF) and monensin while the unfavorable control for the 4C-EQAPOL panel used only dimethylsulfoxide (DMSO) (no Costim) and BRF. The 11C-EQAPOL-1 data were used to demonstrate the difficulties encountered with an endogenous background response where the 11C-EQAPOL-2 data provided a data set with a more common response. All three panels were developed as part of the External Quality Assurance Program Oversight Laboratory (EQAPOL) proficiency program. The lymphocyte subsets for these three data units are available through http://duke.edu/～ccc14/papers/fscore. 2.2 Sample preparation and ICS assay Normal human donors were leukapheresed in accordance with Duke University’s Institutional Review Table and informed consent was obtained prior to sample collection (Jaimes et al. 2011 Sample preparation and staining were performed as previously explained for the 4-color (Jaimes et al. 2011 and 11-color ICS assays (Ottinger et al. 2008 Snyder et al. 2011 2.3 Manual gating Gating for each data set was performed by highly trained operators in accordance with our established standard operating procedure and the process included extensive back-gating to both maximize signal and minimize noise. Desmopressin Uniform gates were applied within each donor. In Section 3.2 the manual gates and thresholds (observe Fig. 2) from two impartial experts were used to infer a range for the value of method. In Fig. 3 manual gates and thresholds from two impartial labs who participated in the EQAPOL 4-color ICS EP1 Program were used. Fig. 2 Choosing a reasonable determined threshold is usually detailed in Section 3.1. The parameters for the positivity thresholding method were optimized in Section 3.2. In Desmopressin Section 4.2 there are a number of clustering algorithms that were applied to discover cytokine subsets. These methods were realized Desmopressin through the use of py-fcm along with the machine learning package scikit-learn (Pedregosa et al. 2011 The parameters were tuned by hand using a basic grid search approach. We also constrained each method to a single best set of parameters that work for all those three stimulations. We provide in the supplemental materials (http://duke.edu/～ccc14/papers/fscore) a description of these methods and all necessary code required to reproduce the results and accompanying physique. 3 Calculation 3.1 F-score as a tool to identify positivity thresholds The F-score or F-measure (van Rijsbergen 1979 is widely used in information retrieval and statistics to measure the accuracy of a test (Jensen et al. 2006 The F-score balances precision and recall. Precision is defined as and recall as Desmopressin gives a larger excess weight to recall; conversely decreasing the value of gives a.