Flow cytometric-based protocols for assessing anti-MT-2 IgG1 reactivity: High-dimensional data handling to define predictors for clinical follow-up of Human T-cell Leukemia virus type-1 infection

The present work provides an innovative methodological approach to assess the anti-HTLV-1 IgG1 reactivity with practical application in clinical laboratory. Serum from non-infected healthy controls (NI) and HTLV-1-infected patients, categorized as asymptomatic (AS), putatively progressing to HTLV-1 associated myelopathy/tropical spastic paraparesis - HAM/TSP (pHAM) or with clinical diagnosis of HAMJSP (HT) were assayed in two-parallel flow cytometry platforms, referred as: Fix and Fix&Perm protocols. Operating-characteristics analysis indicated that a single pair of attributes ("serum dilution/cut-off") for Fix and Fix&Perm protocols presented excellent performance for the diagnosis of HTLV-1 infection. Conversely, Fix and Fix&Perm protocols displayed weak/moderate overall performances when applied with prognosis purposes of HTLV-1 infection. A panoramic snapshot provided by the reactivity boards revealed clearly the higher sensitivity of Fix&Perm protocol for detecting seropositivity for HT, suggesting that stepwise combinatory criteria would improve the global performance of using a single pair of attributes. Three data mining strategies were tested, including endpoint titer analysis, heatmap assemblage and decision tree analysis. Bi-dimensional heatmap analysis demonstrated that, while the clustering profile of NI vs HTLV-1(+) revealed segregation in opposite poles, AS vs HT presented discrete segregation but still displaying an intertwined distribution pattern. The combination of methods for segregating AS from HT displayed a moderate but superior global accuracy (85.7%; LOOCV = 71.4%). The comprehensive data analysis support that the combination of methods have improved the performance to the differential diagnosis of AS and HT, with direct association with laboratorial records, including serum cytoldne levels and proviral load. (C) 2017 Elsevier B.V. All rights reserved.