However, distinct from traditional EMM loads, eplet profiles did not include all possible donor-recipient EMMs related to the HLA locus or class but were composed of only a subset of simultaneously occurring EMMs (informed by donor-recipient pairs in our cohort). and developed models to select the subset of EMM most predictive of DCGF. Results Of 412 EMM observed, 119 class I and 118 class II EMM were associated with DCGF. Network analysis showed that although 210 eplets Rabbit polyclonal to CREB1 formed profiles of 2 to 12 simultaneously occurring EMMs, 202 were singleton EMMs that were not involved in any profile. A variable selection procedure identified 55 single HLA class I and II EMMs in 70% of the dataset; of those, 15 EMMs (9 singleton and 6 involved in profiles) were predictive of DCGF in the remaining dataset. Conclusion Our analysis distinguished increasingly smaller subsets of EMMs associated with increased risk of DCGF. Validation of these EMMs as important predictors of transplant outcomes (in contrast to acceptable EMMs) in datasets with measured allele-level genotypes will support their role as immunodominant EMMs worthy of consideration in organ allocation schemes. value estimates for the EMM associated with DCGF were, at minimum, smaller than the 99th percentile of a random distribution of values estimated under the null hypothesis that the EMM have absolutely no effect on DCGF. To inform the role of eplet frequency on the observed associations with DCGF, we measured eplet distributions in donor and recipient populations. To model the complex relatedness between HLA EMM, we applied weighted correlation network analysis.20 We then evaluated profiles of EMM as risk factors for DCGF by fitting AFT models. A profile was deemed present only if all associated EMMs were observed in the donor-recipient pairs. To identify a subset of EMM significantly associated with DCGF, we also applied Lasso penalized Cox regression21, 22, 23 onto training (70%) and test (30%) datasets. This method enabled feature selection by shrinkage of the number of EMMs among several and potentially correlated EMMs. HRs of DCGF of selected EMMs were then estimated by multivariable Cox regression models while accounting for false discovery rate. A similar selection of EMMs associated with DCGF was identified when including cold ischemia time and donor type (living donor as well as standard criteria and expanded criteria deceased donor). Finally, we conducted sensitivity analysis to confirm the consistency of risk associated with singleton and profiles of EMM in an independent dataset of 48,384 pairs of PRA 0% transplant recipients and their donors. In addition, given concerns that the genotype imputation may be less accurate in non-Caucasian populations,24, 25, 26 we repeated our main analysis in a subgroup of self-reported Caucasian donor-recipient pairs. Statistical analyses were Biotin sulfone performed using the free statistical computing R software (https://www.r-project.org). Results Following application of the exclusion criteria (Figure?1, Study Flow Diagram), a total of 118,313 first-time KTR- (January 1, 2000, and January 1, 2015) from the U.S. SRTR with peak PRA 0% were included in Biotin sulfone the cohort. Baseline characteristics of the cohort and missing covariate data are presented in Table?1. A total of 19,946 KTR experienced graft failure over a median follow-up of 6.39 (interquartile range 3.12C10.01) years. Open in a separate window Figure?1 Study flow diagram. DCGF Risk Associated With Single EMM To evaluate whether AbVer and non-AbVer EMM was associated with DCGF we fit survival models. A total of 449 potential eplets for HLA-A, B, C, DRB1, and DQB1 appeared on the HLA Epitope Registry when accessed in September 2018. Of those, 412 EMMs were observed in the study cohort with 243 EMMs (121 class I: 46 AbVer and 75 non-AbVer, and 122 class II: 48 AbVer and 74 non-AbVer) statistically significantly associated with DCGF in Cox proportional hazards models that considered a single EMM at a time, adjusted for pertinent donor, recipient, and transplant characteristics, and controlled for false discovery rate. Given that the proportionality of hazards assumption was Biotin sulfone violated in many of the fitted Cox models for single EMM, we also fit AFT models (possible distributions of survival times can be found in Supplementary Material S1). Of the 412 EMMs observed in the study cohort, when adjusting for the same variables as the Cox models, the AFT model found 237 (119 class I [44 AbVer and 75 non-AbVer] and 118 class.