High density genetic linkage maps are important tools for QTL fine mapping, map-based cloning, comparative genome analysis and the integration of genetic and physical maps. Several genetic linkage maps based on various markers technologies are now available for perennial ryegrass [1–9]. These maps of moderate marker densities have proved valuable for mapping QTL to broad genome regions. Public marker resources recently established provide the opportunity to increase marker density of these maps, thereby improving map resolution [10–13].
For example, the genetic linkage map of the perennial ryegrass mapping population VrnA has initially been used for a QTL study to characterise vernalization response and contained 93 markers spanning 490.4 cM with an average distance between markers of 5 cM . This map has been complemented over time with candidate gene-based CAPS markers to study disease resistance traits [14, 15] and contained around 180 markers with total map length of 487 cM when used to evaluate seed yield and fertility traits . Recently, the same map has been used to localise genes involved in water stress and contained 222 markers, between 24 and 37 on each linkage group (LG), spanning a total of 736 cM .
Among the different marker technologies available to increase the density of a genetic linkage map, SNPs have attracted much interest, mainly for two reasons: Firstly, SNPs are the most abundant form of genetic variation  and occur at regular intervals in the genome . Secondly, SNPs are highly suitable for multiplexed genotyping assays on mass spectrometry, microarray or beadarray-based platforms . Advancements in these technologies has enabled increased throughput at low cost per data point.
The potential of SNPs for extensive genome analysis has been impressively demonstrated in model plant species such as Arabidopsis thaliana, rice (Oryza sativa), and maize (Zea mays), where fully sequenced genomes resulted in the identification of millions of SNPs suitable for genome-wide association studies and molecular breeding concepts such as genomic selection .
In species where a reference genome sequence has not been established yet, several strategies for large-scale SNP discovery have been reported, mainly being divided into in vitro and in silico approaches. Amplicon resequencing is an in vitro approach and has proven very reliable for SNP identification with a false discovery rate usually below 5% . Furthermore, cloned PCR fragments and allele-specific sequencing allow haplotype identification at sufficient read lengths and the discrimination of orthologous (allelic) and paralogous (derived from closely related genes or highly conserved domains in gene families) sequences. However, amplicon resequencing requires an enormous effort for large scale studies, since each gene needs to be amplified individually and thus might have limited application in the future. Despite the labour intensive nature of amplicon cloning and sequencing, this has been the method of choice for SNP discovery in ryegrasses to date . For in silico SNP discovery, the rapidly growing public EST databases can be exploited as a potential sequence resource [24, 25]. This approach has been applied in other Poaceae crop species including wheat (Triticum aestivum L.)  and barley (Hordeum vulgare L.) . However, availability and quality of public ryegrass EST sequences are often limited and it might be difficult to obtain a sufficient number of EST reads from the same gene, a key factor for reliable in silico SNP identification [22, 28]. As a result of these limitations, the percentage of false discovery rates is often considerably high and can vary between 5 and 50% . Recent advances in NGS opened up the opportunity for whole genome resequencing as an extremely powerful strategy for in silico SNP discovery at appropriate sequence coverage. However, de novo assembly of short NGS reads is difficult in outbreeding species with a highly heterozygous, large and complex genome containing a high degree of repetitive elements. Moreover, whole genome resequencing may not be necessary to target recombination blocks present in bi-parental mapping populations. Therefore, different strategies for complexity reduction such as reduced representation libraries (RRL) have been proposed to sequence only a subset of the genome for SNP discovery . RRLs have been applied in a wide range of plant species such as maize , rice , grapevine species (Vitis spp.) , common bean (Phaseolus vulgaris L.)  and soybean (Glycine max L.) . Another strategy for complexity reduction is transcriptome sequencing [35, 36], where expressed genes are targeted and highly repetitive non-transcribed genomic regions are excluded. This emerged as an efficient method for the high-throughput acquisition of gene-associated SNPs [37, 38].
For SNP genotyping in a scale up to 3,072 SNPs, the Illumina GoldenGate technology  has successfully been used in several crop species. In diploid barley, for example, custom oligo pool assays (OPAs) have been designed to estimate linkage disequilibrium (LD) in inbred elite varieties  and for genetic linkage mapping . Recently, two validated 1,536-SNP barley OPAs (BOPA1 and BOPA2) were made available to the barley community as an excellent marker resource in terms of distribution and density in the barely genome, technical performance and biological importance . In more complex genomes such as soybean, GoldenGate genotyping has been used for linkage mapping in recombinant inbred line mapping populations . While also being autogamous, soybean contains around twice as many gene paralogues (32%) when compared to 16% in barley , which is known to affect the success rate of multiplexed high-throughput genotyping methods [45, 46]. However, the rate of 89% successfully scored SNPs indicated that the genome complexity of soybean had limited impact on GoldenGate performance in a carefully selected SNP panel . In maize, the genome contains about 80% repetitive sequences and a similar amount of paralogous sequences as soybean , but a substantially higher intraspecific genetic variation . Despite this, OPAs containing 1,536 SNPs designed from publicly available SNPs (http://www.panzea.org) are routinely used for diversity, linkage and association analysis, as well as for LD estimations [48, 49]. To date, the GoldenGate assay proved even successful for SNP genotyping in tetraploid and hexaploid wheat lines  and allopolyploid Brassica napus.
Encouraged by this, we developed the first open access Lolium 768-SNP OPA (thereafter referred to as LOPA1) for the allogamous forage grass species L. perenne with a genome size and complexity comparable to maize. Specifically, we aimed at (i) developing an efficient strategy for in silico SNP discovery based on next generation transcriptome sequencing, (ii) implementing a pipeline for successful OPA design, (iii) getting first insights to cross-species amplification rates of ryegrass SNPs and (iv) constructing a high density EST map in perennial ryegrass as a promising tool for QTL fine mapping, map-based cloning and comparative genome analysis.