For large genome outbred species, reducing genome complexity, optimizing barcode and simple procedure are key points for sequence libraries preparation.
The step of selecting fragments appears to be necessary for outbred species for the aim of obtaining sufficient sequencing depth for calling more accurate SNPs and genotyping. Currently, most of available genome reducing and sequencing methods are based on restriction enzymes (REs) 
. The key points for selection of appropriate REs are to avoid frequently cutting repetitive elements and to generate ranges with fragment sizes suitable for sequencing coverage across genome and sequencing depth. The results of several digesting experiments using a few REs showed that Ava
II was a relatively perfect REs for the pig sequencing experiment (data not shown). To compare with GBS method, we selected digesting fragments of ≥200 bp by gel purification to sequence because digesting fragments of <200 bp would be partly or completely sequenced twice or more by paried-end using Hiseq 2000 platform. Although these repeating sequencing fragments improved the accuracy of SNP calling, the method probably increased the variation of reads and decreased the efficient utilization rate of reads. Some next-generation sequencing methods including GBS were excellent to genome-wide genotyping for inbred population with low coverage and small genome animals with low cost. For outbred populations, using these methods to accurately genotype was difficult to achieve with low depth and coverage because of the high extent of heterogeneity and phase ambiguity in the haplotype. The genotyping results of pig experiment showed that GGRS were qualified for outbred population. Furthermore, for other outbred animals with smaller genome, lower heterogeneity or more information than pig (such as chicken), if lower genome coverage and sequencing depth are desired, the procedure can be modified using different restriction enzymes or altering sequencing fragments and thus GGRS pipeline is generally applicable.
Our GGRS procedure employed one set of adaptors. Two ends of digested resulting fragments were ligated to identical barcode-adaptor, instead of one end ligated common-adaptor and the other end ligated barcode-adaptor. This barcode-adaptors design was beneficial for outbreed population by increasing fragments consistency between individuals. The GGRS is appropriate for parallel genotyping of large number of samples than existing RAD protocols because of the simpler procedure, and the optimized protocol for libraries preparation is helpful for saving cost and labor. Moreover, the simplified protocol allows us to use small amounts of DNA for libraries preparation (about 100 ng, even lower) that is important for studying rare materials.
Recently, a streamlined restriction site-associated DNA genotyping method called 2b-RAD has been published 
. The choice of restriction enzyme type of 2b-RAD is a good idea that produced even coverage across genome. In addition, this method can change marker density by modifying the overhang bases of adaptors. However, the flexibility of changing marker density is not enough because of the restriction property of type IIB restriction enzyme. Further, the lengths of the restriction fragments are uniform and short that cannot make full use of the sequencing performance of Hiseq2000 platform with 2×100 bp of paried-end. Compared to 2b-RAD, our methods can generate more abundant data in one lane to decrease cost, and longer reads improve the accuracy of aligning with reference genome although the advantage of 2b-RAD method lies in the well-distributed of markers. Both 2b-RAD and GGRS approaches can perform de novo
analysis for the outbred species lacking an assembled genome sequence easily. In such situation, clustering reads can be regarded as the reference sequence.
Our GGRS pipeline can process 504 samples each run (72 samples/lane ×7 lanes/flow-cell, one lane as control) and >70, 000 pig SNPs can be identified in a short time for an expenditure of $80 (USD)/sample. Up to 288 samples each lane (2016 samples/flow-cell) will possibly be sequenced as along with increasing reads density of upgrading Hiseq 2000 sequencer. These improvements will accelerate the reduction of the genotyping cost to <$20 (USD)/sample.