With promises of higher throughput and lower cost, the new short-read sequencing technologies have garnered a great deal of attention even though their usefulness beyond resequencing is somewhat debatable. This paper presents an interesting twist on hierarchical sequencing and manages to break a complex assembly problem into more manageable chunks.

There are two concerns, however, with the proposed protocol. The first concern will probably go away with time as the process becomes more and more automated. But right now, the handling of 200,000 different fragments and the ligation of different 5-base tag strike me as being highly error-prone and laborious.

The second and more serious concern is whether the proposed approach could remain successful with ultra-short reads like those produced by Solexa or ABI SOLiD. These machines have even higher throughput and lower cost than pyrosequencers, so there is tremendous economic incensitive to use them. But the reads they produce are so short that even SHRAP may fail to generate useful de novo assembly.