Reviewer 1's Review

-----

The authors use the well-established AFLP process to reduce genome complexity. The AFLP products are then made into a library suitable for 454/Roche high throughtput sequencing. This is done on at least two genotypes, followed by bioinformatic extraction of SNPs. Over 80% of such SNPs have been confirmed experimentally.

The paper is well written and describes a useful technique.

One disadvantage of the current implementation is that there is no selection againast repetitive sequences. In comparison with sequencing EST libraries from two different genotypes this method is likely to produce more repetitive sequences (although the Authors select agains such sequences bioinformatically in the postprocessing step). The authors suggest some modifications of their protocol that may reduce the amount of repetitive sequences obtained. It may be good to add some additional discussion of these issues.

Parts of the process involve use of commercial apparatus and possibly proprietary software. Please clarify if the software developed for SNP detection in 454/GS20 reads is available and under what conditions.

Maize may not be the most convincing case for using this method -as it is a very polymorphis species and a large number of public SNPs are already available. Have the Authors attempted to use this approach in another, less polymorhic species? (tomato?)

It would be also worthwhile to mention that SNPs discovered as polyporphic loci between two cultivars or accessions, if used in a larger collection of germplasm, will be burdened by ascertainment bias, and thus not suitable for unbiased discovery of philogenetic relationships.

-----

N.B. These are the general comments made by the reviewer when reviewing this paper in light of which the manuscript was revised. Specific points addressed during revision of the paper are not shown.