Okay, it's become clear that I've let the twittering controversy pretty much take over. But, why stop now? Actually, two additional points on that score (though also chuckling at the sniping at someone talking about cloud computing who wouldn't allow blogging).
First, an argument given for closing some talks is the discussion of clinical data. Some talks were shut down entirely, whereas others asked that only the clinical parts be withheld. This seems like a rather flimsy fig leaf. Any details discussed before an audience of a few hundred people can't be considered very private. As some have commented, at AGBT (and I have certainly seen this at other meetings) folks were photographing away during sessions under some degree of closure. Blogging bans primarily affect the honest and really are little protection from anything being disseminated.
Second, Owen White made a reasonable argument that any ire should be redirected from the speakers to the conference organizers. In his view, some speakers may have felt ambushed and indeed may be unaware of Twitter or how it might be something more than frivolous.
On a more positive front, one speaker actually phoned in his talk due to travel difficulties and apparently did a fine job. Another speaker (Joseph Boland) wore a t-shirt reading "Tweet me!". Curiously though, I don't believe anyone did past noting the shirt!
One thing from an earlier day (Day 1, I think) that I've been mulling is some discussion that exome sequencing seems to periodically miss regions and it isn't clear whether these are truly deletions or just sporadic dropouts. Also, full genome sequencing apparently gives more even coverage of the genome. If this is all true, then the advantages of full genome sequencing over exome sequencing are becoming more pronounced, at a time where the differential in cost is shrinking. Will targeted capture of human exomes be a phase that lasts only a year or two and is then displaced?
On the other hand, one of the talks Anthony Fejes blogged on was work at the Broad using the same capture methodology to pull parasite DNA away from their hosts, in this case Plasmodium (malaria) from human. This can be done with synthetic DNA, or by converting the genome en masse to capture probes.
Another bit of Anthony's notes I mined a bit was a talk on Ion Torrent given by Chad Nusbaum of the Broad. One interesting claim is that the sequences show very little GC bias. I'm left wondering why this plays out this way; is there a trick to their emulsion PCR? Will Life be able to boomerang that back to SOLiD? On the other hand, the quality values are apparently trending on the low side. As I've stated before, Ion Torrent (and other new platform makers) really need to push data out early & often -- they'll win more users by being upfront (we'll find the right applications!) than they'll lose. Also, they'd do themselves a favor by getting more of their customers to map their sequencers; right now there are only two locations with PGMs mapped (though AGBT could be mined to pin a few more).
Ion Torrent apparently also suggested they will launch a new chip every 6 months. If they really can get their 3rd chip out this summer, then they could pull ahead of MiSeq before any of those boxes hit the street. Of course, MiSeq will still have the advantage of the huge ecosystem of Illumina-compatible kits and protocols. Ion Torrent isn't doing themselves any favors on starting to level that playing field; if anyone can find information on how to design amplicon libraries for Ion Torrent on their website they deserve a prize. Ion Torrent also claimed they will get the 8 hour sample prep down to 2 hours; that would certainly be a big difference.
I've skipped over a bunch of other good summaries on Fejes.ca.