Given that EUChip60K consists of a powerful genome-wide-set of SNPs transferable across Eucalyptus kinds (Silva- ), it can enable it to be in depth interspecific studies away from genome design past the individuals said yet (Hudson mais aussi al
The EUChip60K Alt.com ekÅŸi genotyping platform (Silva- ) allowed the construction of the two highest density linkage maps for Eucalyptus (Fig. 1; Table S1). Both maps indicated a few, but clear, assembly inconsistencies in the current genome version, in line with our previous report (Petroli et al., 2012 ), which also pinpointed problems especially on chromosomes 1 and 4 (see Fig. 2 of that report). A recent mapping study highlighted the same facts, and corrected the assembly inconsistencies in a version 2.0 genome available in P hytozome (Bartholome et al., 2014 ). Smoother Marey map profiles for chromosomes 1, 4 and 8 were observed when we aligned our E. grandis map to this new assembly version (Fig. The high marker density provided by our much higher density linkage maps could further help to improve the reference genome in future assembly efforts. Our mapping work also demonstrates that the EUChip60K provides a powerful SNP genotyping platform for any future linkage mapping project. , 2012 ). Furthermore, with > 80% of the SNPs at < 10 kb from 30 444 annotated genes in the Eucalyptus genome, gene discovery from high-resolution quantitative trait locus mapping becomes a true possibility. The E. grandis linkage map reported provided a genome-wide average recombination rate of c. 3.18 cM Mb ?1 , somewhat higher than earlier estimates from DArT mapping (Petroli et al., 2012 ), and a fairly similar recombination pattern across chromosomes with no major recombination deserts as typically evidenced by extensive plateaus of recombination in other plant genomes (Sim et al., 2012 ; Bauer et al., 2013 ; Lee et al., 2013 ).
Recent genome-wide analyses in Populus have shown the average LD declining to r 2 < 0.2 within c. 3–6 kb, substantially more slowly than previous estimates from candidate gene studies (Slavov et al., 2012 ). LD up to 110 kb was also reported when longer sequence regions of Cryptomeria japonica (Moritsuka et al., 2012 ) were surveyed, and earlier reports in Populus (Olson et al., 2010 ), Fagus (Lalague et al., 2014 ) and flagship conifers (Heuertz et al., 2006 ; Eckert et al., 2010 ; Larsson et al., 2013 ) have also indicated somewhat variable extents of LD when SNPs at variable distances were surveyed in longer candidate gene stretches. Our analysis in E. grandis provides solid evidence at a truly genome-wide scale to support a new picture on the extent of LD in an outcrossed undomesticated forest tree genome. By genotyping several thousand SNPs positioned at fairly regularly spaced intervals along the entire genome, across coding and noncoding regions, we were able to capture both moderate and long-range LD. At such a genome-wide scale, r 2 between pairs of SNPs fell to half its initial value within c. 3.7–5.7 kb (Figs 3, S3). LD was, however, quite variable across the genome and pairwise estimates of r 2 spanned the entire range of values, from absence to complete LD even up to 50 kb distances (Fig. 3). No difference was seen in the genome-wide extent of LD decay whether using SNPs filtered at MAF > 0.05 or MAF > 0.01 (Fig. S3), although rare SNPs are known to tend to have lower pairwise r 2 -values (Pritchard & Przeworski, 2001 ). This result suggests, however, that our genome-wide estimates of LD based on a relatively large sample of 72 genomes and over 21 000 SNPs are less sensitive to MAF, at least down to 0.01, than LD estimates in short sequence stretches (Lalague et al., 2014 ), although the true properties of LD around very rare SNPs (MAF < 0.01) still represent an unsettled issue (Pe'er et al., 2006 ).