The main difference from the studies presented in [ 2] and [ 6] on the selection of markers, is the exchange from classification problems (SVM) to regression problems (SVR) as the PTA milk is mapped by a continuous variable.
We therefore estimate that, within any one strain, about 20% of sequences (other than short tandem repeats) are present in multiple copies (11, 12 and 9 out of the 47 markers, in the three strains analyzed), if our selection of markers is representative.
ICIM is an improvement over the commonly implemented composite interval mapping (CIM) algorithm, in which selection of marker is done through a stepwise regression taking into consideration of all markers simultaneously, leaving the flanking markers at the current interval.
This positioning of markers was established by selection for mitotic crossing-over in heterozygous diploids.
In contrast, in genomic selection, a large number of markers is used to predict the performance for complex traits controlled by many QTL with small effects [ 6, 7].
However, for background selection, in which a larger number of markers are needed to provide an overview of the different parental contributions to a progeny's genome, the cost and time needed for microsatellite assays are high or even prohibitive.
After obtaining full array data for 133 isolates, a selection of markers showing variation between individual strains was made.
A selection of marker gels and traces is available as File S2. Results of marker inspections are described below.
Finally, selection for the loss of marker was performed in YPD containing 1 mg/ml 5-fluoro-orotic acid (Thermo Scientific).
Selection of gene markers was done entirely by computational analysis without resorting to the literature and our prediction results were compared with findings of the data contributor.
Selection of these markers was based on their regulatory association with Nuclear Factor-kappaB (NF-kB) [ 41] and their association with IS and PCS in the CKD population [ 42].
