Today, Sven presented an invited talk at JOBIM 2010 in Montpellier about future challenges of next generation sequencing.
I put emphasis on the following key points.
- I agree with David Haussler's HiTSeq talk that it would be interesting not to receive only DNA reads and qualities from the sequencers, but richer information, which we may call "sequences of internal machine state". In a technology-dependent process, they could always be converted to DNA+qualities, but may be more helpful for further analysis.
- I would like to have a lightweight, fast, memory-efficient indexing method that allows to retrieve 1-difference matches of q-grams efficiently. I assume that new ideas in hashing may help here.
- In the past, speed has be more important for read mappers than accuracy. I believe that there is a need for mappers that give stronger guarantees. For example: output not only the best match (or one random best match), but all matches with a certain (quality-weighted) error rate, and then do something which I like to call "probabilistic" mapping, if the location of a read in the reference cannot be determined uniquely.
- With terabytes of sequencing data becoming available, we need to think about large-scale phylogeny algorithms and visual tools for explorative analysis of large multiple alignments.
- In the long run, we will want to develop genrative probabilistic models for pan-genomes (similar to, but richer than HMMs for protein families).
This summarizes it pretty much.
I was awfully tired, because the night before, my flight could not land at Montpellier airport due to bad weather, and so we were re-routed to Marseille, where Air France eventually organized a bus to drive us to Montpellier in the middle of the night. It might have been a nice trait to offer an option to drop us off in the city (who wants to stay at the airport anyway), but no luck. It took another 30+ minutes until a few taxis arrived. By the time I got to my hotel, they had shut down for the night, without an emergency phone number or a night porter available. Fortunately, it was a warm night, so Montpellier is not so bad for homeless people.
In summary, quite an unpleasant experience, and Air France could at least have offered or paid for dinner (as they are obliged to by EU regulations, anyway).