What can you do with 1000 base pair reads?

Has anyone else noticed this yet? The page "Future of 454 Sequencing" on the Roche website poses the rather tantalising question: What can you do with 1000 base reads? The page is accompanied by a rather sexy graph showing a 454 Titanium run acheiving a modal read length of 766 base pairs. For those not acquainted with the current technology: Titanium typically produces 400-500 base pair reads, Solexa between 25-100 depending on run time (typical values 36 and 75) and Solid seems to be run around 50 base pairs at most places. With the short read technologies the quality scores tend to drop off towards the end of each sequence. 454 is no longer considered a short-read technology and in fact 1000 base pairs is getting close to pushing the limit of what skilled users could do with traditional ABI machines using the Sanger technology.

So, what can you do with 1000 base reads? Well, for bacterial genomes this raises a few interesting possibilities. The one that strikes me as most useful will be sequencing almost 2/3 of 16S genes when performing phylogenetic profiling. De novo sequencing will become even easier, and combined with paired-end, the concept of single scaffold assemblies should become reality. I guess it will make whole genome metagenomics easier. I'm not sure how much of a difference it will make to transcriptomics and resequencing, but it definitely won't hurt!

My understanding is that this technology is in beta testing right now in several genome centres so may be available to regular customers very soon.