To get an overview of the deletion process for each family, we computed the distribution of the copy size relative to the reference copy size (Table 2).
Since an accurate detection and estimate of the copy size is a critical point in any study on LTR retrotransposons evolutionary dynamics, we analyzed the distribution of copy size for each family.
First, we expected that the longest copy detected and the reference copy would have the same size because insertions in genomic copies (as compared to the reference copy) were not taken into account in the computation of the copy size.
The Gret1 family, because of a reference copy size of 10,422 bp, and Noble, due to a combination of large reference copy size and high genomic copy number among the Ty1/ copia families in the grape genome, would likely have contributed the greatest to the retroelement DNA content.
Extracted copies whose sequence harbored stretches of N, indicating the presence of unfinished sequences, were then eliminated in order not to bias the estimation of copy size and the classification of the copies (see below for detail of these procedures).
Four categories of repeats dispersed, tandem, palindromic and gene similarity repeats were identified using REPuter [ 51] and manual verification with copy size 30 bp or longer and sequence identity greater than 90%.
For 6 families (Hopi/Osr27, Osr2, Osr8, Osr20, Osr34 and Osr44), the slight difference in size between the longest copy and the reference copy (99.3 to 99.9% of the reference copy size) is due to the homogenization of the two LTRs in the reference copy (see Material and Methods).
Moreover, with this method, we detect only the accumulation of small deletions (smaller than the copy size), and do not take into account larger rearrangements (hence, the small mean deletion size, compared to the high degree of truncation for most copies).
This may be due to the loci having similar repeat copy sizes (36 47 bp) with the exception of locus V27 with a repeat size of 138 bp.
Equation (11) shows that by having multiple compressed copies, the size of the QCS for each integer transform coefficient can likely be reduced.
