2.3 ELISA cut-off estimation
A subset of the sample comprising positive (n=51) and negative (n=114)
sera from symptomatic and asymptomatic hares that tested positive and
negative for MYXV-DNA by qPCR-M000.5R/L, was used to assess the ELISA
cut-off using the R (R Development Core Team, 2008) package
‘OptimalCutpoints’ (López-Ratón et al., 2014). The cut-off was estimated
based on the criteria of a positive predictive value (PPV)
>0.95, to guarantee
that the animals testing positive were truly exposed to MYXV or to an
antigenically related virus.
To further support the estimated cut-off, we performed a finite mixture
analysis of log-transformed RI10 data, to estimate the mean and standard
deviation of the negative population (Peel et al., 2014). Finite mixture
models allow to characterize the distributions of subgroups within
bimodal datasets (Benaglia et al., 2009), thus being an alternative tool
to estimate the cut-off of serological tests in the absence of reference
tests (Peel et al., 2014). Finite mixture models were implemented using
the expectation–maximization algorithm for mixtures of univariate
normal distributions, from the package ‘mixtools’ (Benaglia et al.,
2009) in R (R Development Core Team, 2017). The proportion of the
seronegative subgroup was set at 1-prevalence and the mean of each
subgroup was defined from the mean logRI10 of the subgroup in which
molecular testing was performed.