It has been a few days since I have returned from Vienna where I attended ISMB07 – this year a joint meeting with the european ECCB series. With about 200 talks and 1000 posters, far too much material to cover it in a blog post. Here a just a few random impressions:
Vienna is a beautiful city (if you like the old European capital architecture with plenty of churches and other large representative buildings with lots of gold applications). The weather was far too hot for my taste, but the conference center and probably most hotels have good air conditioning. The conference center itself was quite pleasant, with plenty of space, decent coffee, and unobtrusive badge controls. I must say that I cannot stand these army-type bouncers with their intimidating “you wear your badge and I won’t hurt you” attitude. This was the impression I had at several previous ISMBs. I heard complaints that the conference center’s floor plan was confusing and that certain poster exhibitions were hard to find. This is probably true, and there also was a clear tendency that poster sections close to the coffee had more visitors than those in far away places. I am not sure how this can be avoided – unless you have these hangar-type poster halls that can easily accommodate 1000 posters in a single room.
Façade in Vienna. Image by Kay at Suicyte
When observing the changes in session layout during previous ISMB conferences, it seems that ISMB is on its way to a more biological and a more ‘normal’ meeting. For me, ‘normal’ is the way that meetings in my area tend to be organized – as these are bioscience meetings, it is well possible that the two changes boil down to one: ISMB on its way to biology. However, there is still a long way to go. In the old days, ISMB talks were recruited from submissions of full papers, which – upon acceptance – were published in the conference proceedings. I have been told that this kind of meeting is customary in the computing sciences. This may be true, but the approach is certainly not viable in the life sciences. No biologists in their right mind would send their best work to a conference (something like nine months before the meeting!) with the sole prospect to be elected as a conference speaker and see your results published at an obscure place that can only be seen by other meeting participants. Over the years, the conference proceedings became published as special issues in Bioinformatics, which at least is a proper journal accessible also for non-participants. However, I am not sure if the conference papers are indexed and can be searched by the normal search pages (of the journal, or in Pubmed). Few years ago, ISMB introduced the PLoS track, fostering contributions with more relevance to the biosciences. As far as I understand, these contributions do not go through the usual ISMB peer review process (with rejection rates above 80%) but are refereed and selected by a different process.
Now, ISMB has another track called ‘Highlight Track’, asking for contributions that have already been published within the last year. Initially, I thought that this track is a perversion of the meeting system, as in the biosciences you expect people to present new work, which has not been published before. While even at bioscience meetings, people often tend to present old work when they are expected to present new stuff, nobody likes this practice and repeat offenders risk not being invited to meetings anymore. At the end of ISMB07, it turned out that I have spent most of the time in these highlight sessions. Here are the reasons why: the highlight talks tended to be more biological, they were typically given by more experienced scientist with better presentation skills, rather than young grad students, and tended to tackle real problems of broad interest. In case that you, dear reader, are a computer scientist, please forgive my ignorance. This was my 8th or 9th ISMB, and in the past years I tried hard to understand how mathematicians and computer scientists tick. Unfortunately, I never succeeded. I still fail so recognize the beauty of a new new algorithm with highly desirable scaling properties, as long as it is applied to biological problem that I consider boring or even moot (e.g. trying to find the provable ‘best’ solution for something, when there already is an abundance of solutions that are good enough). Call me an ignorant, I will not argue. But don’t dare to do, what one (anonymous) lunchtime neighbor of mine did: claim that ISMB is not only attended by biologists but also by some scientists. (Btw, the latter class was meant to comprise mainly mathematicians)
There are still big differences between ISMB organization and typical bioscience conferences. If you want a slot in the regular program, you still have to submit your whole work (including results, discussion etc) many months before the meeting. At a normal meeting ( I am tempted to say “at a real scientific meeting”) , you are either invited (meaning that you don’t have to submit anything – the organizers trust you to show something interesting) or you have to submit an abstract that will be used for the selection. The advantage is that you are able to present results obtained very recently. For example, the results presented in my own meeting talks are typically not older than two or three months, and on at least one occasion I included a piece of evidence mailed to me by a collaborator while the meeting was already underway. Currently, ISMB strongly discourages the presentation of timely results. On the other hand, the ISMB system also has its advantages: the selection of speakers is less biased and leads to a higher degree of diversity. The normal ‘invitation system’ used in bioscience meetings means that you always see the same faces on the speakers podium. For obvious reasons, the conference organizers invite the big shots of the field plus a small number of young scientists who can be expected to present good data (often because they were postdocs in a big shot lab before they started their own group recently). The selection on the basis of submitted abstracts has a similar problem: as not much information is conveyed by the abstracts, it is tempting to select on the basis of the submitters name or working place. Summarizing, I would say that the ISMB way is more fair for the presenters, while the bioscience way is more agreeable for the listener.
At ISMBs, the keynote speakers are the only truly invited people (and supposedly the only ones who don’t have to pay for their attendance). As in other meetings, keynote speakers are selected for their ability to give entertaining talks. With very few exceptions, they come from the field of biology (maybe even the program committee isn’t sure about the idea of a mathematician giving entertaining talks – at least in an area relevant to the ISMB folks). Judging by this year’s experience, the ISMB seems to emphasize the entertainment component even more than other meetings. I have heard several people complain about a lack of science (and substance) in most of the 2007 keynote talks, which were mostly gravitating towards historical overviews with some interspersed jokes. N.b. that this does not apply to all of this years keynotes. There was also the open-access evangelist Michael Eisen (talking about real science), the RNA evangelist John Mattick (talking about how RNA rules the world and proteins are just an semi-obsolete remnant from prokaryotic times) and several others.
Hot topics and trends
As I had skipped the last three ISMB meetings, I expected to see major changes in the subjects covered in comparison to 2003. Surprisingly, this did not really happen, or maybe I just didn’t notice. In former times, there were several waves of hot topics sweeping across the field of bioinformatics and strongly dominating the ISMB talks. Some examples are neuronal networks (being applied to almost everything in the early ISMB days), Hidden Markov Models (applied to sequence analysis, gene finding and lots of other topics somewhat later) and Stochastic context-free grammars (applied to RNA topics and others). I could not make out such a dominating trend this year. In terms of methods, NNs, HMMs, and several others seem to coexist peacefully, sometimes even within the same research groups. If I would have to describe one new development, I would mention a certain trend towards extracting large data samples from literature or databases (e.g. Pubmed or GEO) and then analyzing them in a very generic way that does not require a deep knowledge of the data. I am talking about things like finding co-occurring MESH terms in anonymous text bodies, or finding genes that are co-expressed over large multi-centric data repositories without wasting any thoughts on what these studies were about. This approach is related to what Google does, and seems to be successful if the data bodies are big enough. It also bears some resemblance to the “wisdom of the crowds” approach. The idea is that if you poll the opinion of a vast number of people, you don’t need experts anymore. Or, if you compare the behaviour of some genes over a huge body of random experiments, you no longer need any good experiments. I am not really a friend of this approach, but I must admit that the talks on these subjects were not too bad.
I did a brief survey to see which other blogs talk about ISMB. Here is a random sampling:
- Research Remix here and here and here
- Nodalpoint: here and here
- The Beaver: here and here and here
- Sjoemac: here and here
For a meeting as big as this one (and a large number of bloggers either supposed to be there or actually spotted) is seems like ISMB07 has not made a big impression of bioinformatics blogosphere.
Updates: In the meantime, some more ISMB-related posts appeared: