Results of the present work suggested that the partial least squares regression imputation method could be useful to impute SNP genotypes when pedigree information is not available.
For missing data (partial and complete), statistical imputation with "last observation carried forward" will be used.
Patient numbers and proportions per diagnosis category are expressed as a percentage of the number of patients with full or partial responses, with no imputation for missing values.
In the single-breed approach, the accuracy of imputation using partial least squares regression was around 90 and 94% for the 3K and 7K platforms, respectively; corresponding accuracies obtained with Beagle were around 85%and90%0%.
Note that different prediction models are fitted and evaluated for the following approaches: (i) full-data, (ii) complete-cases, (iii) multiple imputation, and (iv) partial prediction models.
In the multiple imputation analysis all participants with partial information on resource usage and outcomes will be included.
QTL mapping was conducted using interval mapping (using the imputation algorithm to account for partial RIL genotypes) as implemented in rQTL (Broman et al. 2003) (see File S2 for additional details).
The other supernetwork method we consider, Q-imputation [ 35], also uses partial trees as input but uses an alternative approach to generating complete splits that is based on the four-taxon subtrees (quartets) of the partial trees.
We performed genotype imputation using the NIPALS (nonlinear iterative partial least squares, [ 32]) algorithm, which performs principal component analysis in the presence of missing data.
We refer to this sample of 6406 as the 'imputed sample', and we used a data imputation routine to deal with the partial missingness present in these data (more details can be found below).
