A modeling approach for mean fluorescence intensity value harmonization and cutoff prediction for luminex single antigen bead assays of two different vendors

Karahan, G.E.; Haasnoot, G.W.; Voogt-Bakker, K.; Claas, F.H.J.; Roelen, D.; Heidt, S.

doi:doi:10.1111/tan.15082

Luminex single antigen bead (SAB) kits from One Lambda (OL) and Lifecodes (LC) are widely used for HLA antibody detection but have substantial differences in design and assay protocol resulting in different mean fluorescence intensity (MFI) values. Here, we present a non-linear modeling approach to accurately convert MFI values between two vendors and to establish user-independent MFI cutoffs when analyzing big datasets. HLA antibody data from a total of 47 EDTA-treated sera tested using both OL and LC SAB kits were analyzed. MFI comparisons were made for the common 84 HLA class I and 63 class II beads. In the exploration set (n = 24), a non-linear hyperbola model on raw MFI corrected by locus-specific highest self MFI subtraction yielded the highest correlation (class I r(2): 0.946, class II r(2): 0.898). Performance of the model was verified in an independent validation set (n = 12) (class I r(2): 0.952, class II r(2): 0.911). Furthermore, in an independent cohort of post...Show moreLuminex single antigen bead (SAB) kits from One Lambda (OL) and Lifecodes (LC) are widely used for HLA antibody detection but have substantial differences in design and assay protocol resulting in different mean fluorescence intensity (MFI) values. Here, we present a non-linear modeling approach to accurately convert MFI values between two vendors and to establish user-independent MFI cutoffs when analyzing big datasets. HLA antibody data from a total of 47 EDTA-treated sera tested using both OL and LC SAB kits were analyzed. MFI comparisons were made for the common 84 HLA class I and 63 class II beads. In the exploration set (n = 24), a non-linear hyperbola model on raw MFI corrected by locus-specific highest self MFI subtraction yielded the highest correlation (class I r(2): 0.946, class II r(2): 0.898). Performance of the model was verified in an independent validation set (n = 12) (class I r(2): 0.952, class II r(2): 0.911). Furthermore, in an independent cohort of post-transplant serum samples (n = 11) using the vendor-specific MFI cutoffs dictated by the current model, we found 94% accuracy in bead-specific reactivity assignments by the two vendors. We recommend using the non-linear hyperbola modeling approach with self HLA correction and locus-specific analyzes to harmonize MFI values between two vendors in particular research datasets. As there are considerable variations between the two assays, using MFI conversion for individual patient samples is not recommended.Show less

Leiden University Scholarly Publications

View statistics

Documents

In Collections

A modeling approach for mean fluorescence intensity value harmonization and cutoff prediction for luminex single antigen bead assays of two different vendors