Sign in or Join FriendFeed
FriendFeed is the easiest way to share online. Learn more »
ISMB 2008

ISMB 2008

For anyone interested in following ISMB 2008, Toronto. Arrange meet-ups, micro-blog the talks and whatever else you like.
Michael Kuhn
This year, ISMB will auto-generate posts for each talk around 10 minutes before the start of the talk, so that the live-blogs can be linked to the ISMB homepage. I hope we'll have lots of coverage! - Michael Kuhn
ISMB will also advertise this on their homepage and in the conference newsletter, so hopefully we'll get many new contributors. - Michael Kuhn
Will we be able to micro-blog the SIGs from this room as well? - Allyson Lister
Yes, no problem. We will generate posts for each SIG meeting. - Tina
@Bettina - thanks :) - Allyson Lister
This doesn't look like a room - it's actually a "friend", even though it's a conference "friend". Is this because you don't want people starting topics there? As it stands, we'll only be able to comment on existing topics, and not post interesting links etc EXCEPT as they relate to an already-started topic. Perhaps this is deliberate to keep things on-topic. Is this what you were going for? Just curious... :) - Allyson Lister
you're right, I didn't realize this. I brought it to the attn of the admins. It might be deliberate (because of the integration with the ISMB website), but then one should probably create another room to coordinate meet-ups etc. - Michael Kuhn
Thanks, Michael. Yes, the idea was to keep it more organized. We will create another room for the "ismbeccb2009-talk" where everyone is welcome to post but please use the official "ismbeccb2009" conference room for all discussions regarding the presentations (Highlights, Proceedings Track, Technology Track, SIGs and Keynotes). - Tina
@Bettina - thanks for the reply. While I'm not 100% sure that splitting things up like this will work (people might not switch that much between the two), I can see the reasoning and hope it works well. Perhaps these reasons should be added to the taglines of the ISMB "friend" and the ISMB talk group you mentioned, so that when people go to one, they see both what it's for and how to get to the other one... Thanks! :) - Allyson Lister
Ok, its true, switching between groups might not work. I talked with my supervisor and we decided now to change the settings. So now everyone of you should be able to post in the ismbeccb2009 room. - Tina
@Bettina - thanks, looks good. Looking forward to the start of the conference on Saturday! :) - Allyson Lister
Michael Kuhn
After talking with Reinhard Schneider, one of the ISMB09 organizers, I opened a new room to try to begin the discussion how to integrate the ISMB homepage with FF/microblogging in general. I hope we can find some low-tech solutions with the help of the FF API and widgets. - http://friendfeed.com/rooms...
Roland Krause
OK, this open doc could be used to collect ideas for the report. If you want to contribute, please leave a message here, preferably with what you would like to add. For now, the document only contains some scribbles and nothing of general interest. - Roland Krause
I have now added the first notes on the literature mining results to the document. I will try to make a figure out of it tomorrow - just a simple plot showing the growth of the terms that I have found. - Lars Juhl Jensen
I've added a draft of my writeup for the Web 2.0 BoF. Feel free to edit! - Shirley Wu
Just to make sure everyone notice: I have gone through out manuscript and made a few comments. I think the manuscript is in a quite good shape, so we should be able to finish it quickly once we are all over the NAR database deadline on September 15 :-) - Lars Juhl Jensen
I added a (draft) small report on the BioPathways SIG. I can trim it down to fit as a box or maybe it can just go in supplementary I don't really now. I will check the main article tomorrow. - Pedro Beltrao
Added few subtle changes to FriendFeed collage. - Daniel Jurczak
Can I get access to it? I am on the Short Read way. - Steve Wu
Shirley Wu
ISMB 2008 = highest profit/participant, as mentioned proudly by the chairman. He assures it's not because of the food, that they are actually spending quite a bit on the food. Maybe it's being delivered to the wrong place.
What I heard is that the number of paper submissions was down by 30%. - pn
can anyone tell me why the espresso machine at the convention center breaks down every moment precisely when i need it most? - Shannon McWeeney
Paulo: So is the amount of free t-shirts. Conincidence ? ;) - Daniel Jurczak
we paid $435/head for food and coffee, the conference center+AV+MCC related costs amounted to another about $180/head. you do the math how much of your money went into food. totally with you on your surprise, but it is actually the same for most places in the world we went to; some were much worse than toronto, others much better (e.g. vienna). - espresso machine: another one of those things: i for one need good espresso: you guys have no clue what the MCC would have charged for 'second cup' quality .. - Burkhard Rost
In Detroit there was no included food and in some days there was almost no place close to the GM Centre with good food (if you call Detroit's greektown a place that you can get decent food ...) and the registration price was adequate. Why include food in the registration when some people cannot afford it? Like me, unsupported post-doc living in Toronto. No way I am shelling 775 (700 reg + annual fee) plus whatever the BOSC costed and eat moderately bad food. - pn
The ISMB is a community effort and the organizers thereby responsible to the community to organize a financially viable meeting. I am glad that Burkhard is as bold as to discuss this issues in public. Organizing a conference is a multi-optimization problem that should have higher weights on scientific content, adequate conference rooms and WLAN (in that order) than food quality, which were met at this years' ISMB for sure. - Roland Krause
Some conferences do not provide lunch at all (e.g. the genetics congress in Berlin in July), but for Stockholm, this is probably not an option (fair ground). - Roland Krause
IMO, good conferences do make a profit, always tough for scientific ones. But the logistics and experience better be worth it then. - Deepak Singh
no lunch is what i have been proposing since years; will become stronger on this! for Stockholm: too late (and yes too difficult); for Boston 2010: we already signed the contract including food, however, we will try our best to get out of it; for 2011 in vienna: difficult - upside: that's perceived as "good food" - Burkhard Rost
profit: half of the ISCB annual budget originates from the ISMB profit, we need $180K from ISMB to survive; we now run ISMB at an incredibly ambitious low budget saving wherever we can, this is one reason why Toronto will create high profit/participant - mostly due to the fact that we cut on organizing costs. still the major issue remains that we pay for the facilities that could welcome 1000 more participants and with every new participant we would profit more. put differently: we have to grow or to shrink - Burkhard Rost
Was that "paper submissions down by 30%" true? $550 - 900 per person. 1000 attendee. Let's say... 30% on foods: $300 X 1000 = $300,000 for lunch and matrix party. Foods must be delivered to the wrong place. - Kuan-Ting Lin
see our paper in PLoS CB on ISMB 2008 about numbers, although the following may not be in there: vienna proceedings submission=492, toronto: 292; ok; NOW please realize that this has absolutely NOTHING to do with costs! One issue: we had over 180 talks at ISMB 2008; 48 were from proceedings=less than 1/3. now the money: food/head was over $400 NOT 30%!!! there were over 1400 participants and the food was way more than 2* your estimate; matrix party was cheap! food & rent for Liberty Grand was a lot. - Burkhard Rost
Hey folks, ISCB almost died from less profitable meetings in Brazil and Australia...give Burkhard a break. I like his openness. Yes, the food could have been better (the box lunches at the Member's meeting during lunchtime were really good!), but I didn't see enough affordable restaurants on Front Street or near the CN Tower to handle us all. I do agree that when possible, letting us... more... - John Greene
In Vienna (that is if it is again at the Vienna International Center) it is not that easy to get to food. I mean there is nothing in the immediate vicinity. Additionally as Burkhard already mentioned the food there was actually quite fine. Anyway, i appreciate the openness of the discussion here. - Daniel Jurczak
Roland Krause
ISMB 2008 report by the organizers (PLoS Computational Biology) - http://www.ploscompbiol.org/article...
Aye. - Roland Krause
note the date: was written BEFORE ISMB, i for one am not good enough to devine "reviews" - we invite you guys to write reports about the meeting and send them to BJ to coordinate where they will go - Burkhard Rost
Pedro Beltrao
"ISCB is looking for a few aspiring reporters, who also happen to be attending ISMB 2008 in Toronto. (...)This year we are also seeking a select corps of students and Jr. Scientists to take on the role of writing about the vast scientific content of the conference and the personal experience of participating in this large, international bioinformatics event. Submissions will be reviewed for posting to the ISMB website, and will be considered for publication in the designated ISCB pages of PLoS Computational Biology, the Society's official journal, as a single author report or as part of a larger report compiled from the contributions of multiple authors." - Pedro Beltrao from Bookmarklet
Hmm, shouldn't we (the ususal suspects) grab the bull by horns and write a decent conference report, compiled from multiple sources? - Roland Krause
folks: another one of those: spent a lot of energy in vain hoping to get something similar to what you guys pulled off: totally impressed!!! Thanks. write it up, we'll put it onto our journal pages if we'd find it ok. other than that: we'll like a blog at a central place in stockholm and will link to this one as soon as steven leard is back from his deserved break. burkhard - Burkhard Rost
Hi Burkhard , I am glad you find this experiment interesting. It ended up being a useful way not only to gather notes about the conference but to set up events on the fly. It would be great to have something like this set up from the beginning at the next ISMB. - Pedro Beltrao
Just add my voice to the fact that I think this was really impressive grass roots reporting. I felt like I was actually at two meetings at the same time, one where I was physically present, and one where I was just a fly on the wall. Great example of what can be done (with enough people and the right tools) - Cameron Neylon
again: have tried to find you guys before; now that i found you i will do whatever i can to support this for future ISMBs; let me give you another impression: i saw the blog for the first time during the wed morning keynote, after the keynote i stood up and thanked Lars & Neil (those were the ones i had seen in the 20 secs i spent on this); at least 30 people walked up to me after that announcement and asked how to get at that blog! - Burkhard Rost
Thanks Burkhard! I think that it would be a great idea to improve the visibility of the live blogging from the official ISMB web site. However, it is also important that we don't try to formalize it too much as this would likely kill a grassroot effort like this. I find that FriendFeed worked remarkably well (much better than my personal attempts on Tumblr), so I suggest that we try to stick to that but somehow tie it in with the ISMB web site. - Lars Juhl Jensen
One possibility would be to set up an ISMB2009 room ahead of the conference and prefill it with one thread per presentation. These can then be used to comment on each talk. By making them ahead of time, we could link to the thread in an iframe on an official ISMB page with the speakers abstract. This would allow people to work in an almost unchanged manner, yet it would structure the comments slightly better and hook it up with the official web site. - Lars Juhl Jensen
Another wild idea: virtual poster sessions. Could we somehow get people to upload a high-resolution bitmap of their poster? These could then again be linked with a FriendFeed thread to allow commenting. Perhaps the posters could even go on Flickr and be fed directly into FriendFeed? - Lars Juhl Jensen
Roland & Neil, I'm also up for helping out on writing a conference report. Burkhard, where would the conference report be published - in PLoS Comp Biol or Bioinformatics? - Lars Juhl Jensen
Lars, we are in touch with BJ Morrison. There is a Googledoc with 5 minutes worth of scribbles by me but I wonder whether a fresh start - or sketch - wouldn't be advantageous. 2000 - 5000 words, I guess, the rest is up to us. Oh and yes, PLoS Comp Bio. - Roland Krause
Roland, excellent - can you email me an invitation to join the Googledoc? I can try to take a look at it later today. - Lars Juhl Jensen
neil: you did not see this because you had obviously left already; i saw your feed first on wed. morning; i stepped up after hannah's talk and announced it and was bombarded by questions: 'how to find that' for the rest of the day, literally at least 20 people approached me. Will anybody write for journals? - Burkhard Rost
Kuan-Ting Lin
ISMB 2008 Conference Proceedings now available on-line at: http://bioinformatics.oxfordjo... - http://www.iscb.org/ismb200...
Pardon me. I don't get it. Where's the proceeding? - Kuan-Ting Lin
Thanks - Kuan-Ting Lin
Michael Kuhn
FYI: proposed to FF to place a comment link at the bottom, and other things we missed at ISMB - http://groups.google.com/group...
Michael Kuhn
Tuesday's circle of laptops - http://friendfeed.com/e...
ISMB 2008 live-blogger's meeting
ISMB 2008
the images are rather noisy, so only low-res versions :( - Michael Kuhn
Daniel Jurczak
Does anyone know, whether the live video feeds of each presentation they are streaming to the iMacs next to the Biomed Central stand are recorded in any way ?
Went to the information desk and asked. People there were surprised to hear that there even is a videocast. So even though, from a technical point of view, everything is in place, there is no recording. - Daniel Jurczak
ISMB sent me a fax to sign to allow them to record my talk and to make derivatives such as videos that would enhance a publication. My talk is for the PLoS track so I assumed that it was for them to record the talks to add to SciVee. - Pedro Beltrao
1) glad you liked it; was my idea and a hell of work; 2) no recording because (i) everyone else felt this is too low quality, (ii) actually, not that simple to record what is streamed is it? - Burkhard Rost
Pedro Beltrao
some pictures from ISMB2008 (sorry for the duplication) - Pedro Beltrao
want to put up a web site with many photos from ISMB 2008: anyone to help? - Burkhard Rost
Shirley Wu
We're putting together a conference report but would like to include the more "human" side of things. If you have funny stories, anecdotes, speaker quotes, reflections, or would like to contribute, just encapsulate it in 1 para or less and we'll fit in as many as we can. Paste here or link to it.
great: let ISCB know, we'll link to it! - Burkhard Rost
Michael Kuhn
Conference hack: Embracing the backchannel at Start - http://www.veen.com/jeff...
the Start conference will have an ombudsman following twitter/IM/email to give the audience a voice. nice idea. - Michael Kuhn
Shannon McWeeney
HL 24: eQED
Initial Example Yeast QTL strains - Shannon McWeeney
"Genetical Genomics" integration and analysis of genotyping and expression to identify regulators - Shannon McWeeney
Reviews common problems: large genetic regions, multiple hypothesis testing, and that complex trait composition applies to regulators as well (multiple regulators potentially of small effect and with redundancy) - Shannon McWeeney
Integration of physical interaction data - with idea that it will give us true regulators. If we have 10 genes in locus, true regulator should be closer wrt network distance than others to target gene. random distribution of others (not sure if i agree with that- could be other types of correlation) - Shannon McWeeney
Need scoring to rank interactions and identify true regulator - Shannon McWeeney
Simulation (based on idea of communication as electric current). weak confidence equivalent to high resistance etc. - Shannon McWeeney
Note: score all pathways not just most likely - Shannon McWeeney
Computational Fine mapping example. reference was knock out expression data in yeast. 550 associations between regulator and target gene. random accuracy =22%. Tu et al random walk 48%, shortest path 64% eqed 72% - Shannon McWeeney
Integrating several e-qtl loci. Idea similar to epistasis that main effects may not be significant but joint effects are important. Inclusion of genes not significant but which connect loci - Shannon McWeeney
Important point: several e-qtls that reinforce same part of the network. ability to mine integrated information - Shannon McWeeney
Q: wouldn't bayesian networks be more natural approach rather than circuit with constraints - Shannon McWeeney
Q:how were weights define? Prob based on confidence of PPI. Also ability to exploit correlation in expression - Shannon McWeeney
Pedro Beltrao
I am leaving soon, thanks to everyone that participated. It was a fun and interesting experiment. See you "around"
You're flying out today? - pn
yeap, around 6pm today - Pedro Beltrao
Too soon, we barely talked. Have a safe trip back to California. - pn
I am also on my way to Pearson, it was nice meeting you all (Pedro, Shirley, Neil, Roland..). - Daniel Jurczak
Nice to meet you all as well, it looks like I missed some good sessions/keynotes, but the FF coverage was great! - Shirley Wu
It was great to meet you all! I missed all the talks, but thanks to the FF coverage, I at least have a pretty good idea of what I missed ;-) - Lars Juhl Jensen
Roland Krause
SS07: The future of scientific publishing - The publishers view
Uh oh, no slides "unstructured digital publishing" - Roland Krause
Matt Cockerill (BioMed Central), Claire Bird (Oxford University Press), Catherine Nancarrow (Public Library of Science) - Michael Kuhn
PLoS, BMC and OUP - Roland Krause
MC: quality control function of scientific publishing; peer-review can be a rate-limiting resource for sharing scientific knowledge; need efficient process; will see more sharing of reviews between journals - Michael Kuhn
MC: reviews should separate soundness from how exciting the research is - Michael Kuhn
Validation processes - Roland Krause
MC: medical community more open to open peer review than biologists, might change (Biology Direct) - Michael Kuhn
Open peer review - the Matt Cockerill mentions medical journals and Biology Direct - Roland Krause
Matt Cockerill talked about re-using reviews from higher impact journals to lower impact journals, talked about lower tier (PLoS ONE and BMC Notes), open peer review (works in medicine and Biology Direct) - Pedro Beltrao
much more information structured, needs open access - Pedro Beltrao
Fast talkers and slow internet connections - Roland Krause
MC: would be nice to have full text so that more meta-analysis can be performed - Michael Kuhn
CB: expects to see new networks and tools that can complement impact factors - Michael Kuhn
Claire Bird ... quality of data and re-usability of the article is important and integrity of the publishing process needs to be sustained. community will decide the limits of usefulness of a publication. - Pedro Beltrao
Catherine Nancarrow - alternative peer review model ? self-organizing community online. How comments , ranking etc could be counted for career progress - Pedro Beltrao
CB: is there an alternative to the current model? Discussion recently: self-organizing communities on the web [hmm, didn't Nature try this?] - Michael Kuhn
Question by Pedro: We can't find out currently how many people download a paper. MC: Publishers recognize this, think about how to share access data; perhaps fully open citation base [Thomson won't like this] - Michael Kuhn
Q: Cooperativity and data-sharing – author lists and order is an incentive against data sharing; "support" scientists don't get the credit they deserve; perhaps author list in alphabetical order would be more egalitarian. CN: Require author contributions in the meta-data. Strive to be as transparent as possible. MC: With the right collection of data, these "hidden" contributions could be measured - Michael Kuhn
Peter Karp - maybe we could have an ontology of contributions (sounds interesting) - Pedro Beltrao
Q: Perhaps we could have an ontology of contributions? Or the author contr. could be aggregated - Michael Kuhn
Reactome developer - it has been very difficult to have Reactome tracked in Pubmed. - Pedro Beltrao
Q: someone from Reactome; complains that it took a long time to have the database listed in PubMed; is there a way to make this more visible [didn't fully understand this] MC: Interested to talk to UniProt guys, current tools are drived towards journal articles, but they should also track DBs - Michael Kuhn
Discussion about Open Access (not really about publishing on well) - Pedro Beltrao
MC: NIH repository policy might lead to freely accessible publications, which are still under restrictive licenses (re: text-mining) - Michael Kuhn
Roland Krause
SS07: The future of scientific publishing - G. Cesarini on Structured Digital Abstracts
A project underway at FEBS Letters in collaboration with the MINT data base - Roland Krause
Large gap between articles on PPI and publications of low throughput experiments - Roland Krause
started working in structured information due to experimental work going on in the lab. Explained the creation of MINT and the iMEX consortium. The annotation by curators are not covering no-where near enough the increasing number of pubmed articles - Pedro Beltrao
The principle of structured digital abstracts (SDA) http://www.febsletters.org/article... - Roland Krause
Project started in January, first issue (see above link) in April 2008 - Roland Krause
Link to press release with useful information http://www.eurekalert.org/pub_rel... - Pedro Beltrao
The authors can reject doing the structured abstract, I think this is where it might fail, if most people don't want to do it.3 in 4 accepted .. not bad, only 30 structured abstracts so far. 16 out of 30 authors filled a survey about the process.Some points: 1h of average time to do it. 6/10 had some difficulty. the most difficult task was the find the uniprot ID (wow). 2/16 said that this could affect their decision to publish there again - Pedro Beltrao
161 articles published, 30 with SDA - Roland Krause
90% wrong assignment on taxonomy - Pedro Beltrao
The data is almost insignificant in relation to the whole field but the information int he FEBS experiment will be useful for publishing world. - Pedro Beltrao
The general problem so far has been that authors make mistake so curators still have to spend considerable time on the process. Pubmed did not accept to add the structured abstract to the Pubmed abstract. - Pedro Beltrao
in the future they hope to have more journals participating and more types of information captured (they focus on articles that describe protein-protein interactions) - Pedro Beltrao
Roland Krause
SS07: The future of scientific publishing - Mark Gerstein on text mining to study the structure of science
Introduction: the rapid growth of databases and publications - Roland Krause
Blurring the distinction of databases and journals, particular in biology as a fact based science - Roland Krause
distinctions between databases and journals are blurring - Michael Kuhn
science 2.0 == science about science (??? wouldn't that be meta-science ???) - Michael Kuhn
New Vision: Science 2.0: studying the structure of publishing - Roland Krause
Science 2.0 as the science of science ( I don't think I agree with this definition) - Pedro Beltrao
Seeking a New Biology through Text Mining, Andrey Rzhetsky; http://www.cell.com/content... - Michael Kuhn
Web mining: extract simple statements, find new information. - Roland Krause
Structure of science: Co-authorships, subjects - Roland Krause
Seringhaus et al., Chemistry Nobel Rich in Structure, http://www.sciencemag.org/cgi... - Michael Kuhn
Mashups: Search for nobel prize by subject http://www.ncbi.nlm.nih.gov/pubmed... - Roland Krause
using text mining to understand the structure of science, how ideas evolve, etc (the net is not great here). highly studied genes, knowledge bias. I think Valencia has been one of the first to show this. - Pedro Beltrao
Would also be easier to follow a less productive scientist - Roland Krause
Uncovering trends in gene naming; Michael R Seringhaus et al; http://genomebiology.com/2008... - Michael Kuhn
comparing literature nane over-representation with something like google search rank for gene names. Not surprisingly the names that are more abundant in "human discourse" than in literature are "funny" names that already exist. - Pedro Beltrao
comparing name co-authorship graphs in different institutes (that is an interesting idea .. it would be interesting to do this for different science topics) - Pedro Beltrao
Structure of co-authorship networks in relation to consortia to identify bias in selection of targets for structural genomics projects - Roland Krause
mapping science topics as whole (I did not get the name of the authors for this) - Pedro Beltrao
Giving the example of RNAi where the birth of the field is captured in the databases. Clusters of co-authorship graphs over time - Pedro Beltrao
The impediments to the vision of a connected way of scientific publishing - Roland Krause
Federation rather the centralization - Roland Krause
getting to the idea of structured abstract. Involving the authors in the production of a machine readable abstract for the work.Saying that even a small set of structured abstracts could be used as gold standard. - Pedro Beltrao
Notes the absence of a social framework for protecting data on the web - Roland Krause
social framework for the protection of papers and data on the web. We don't know how to regulate data sharing online. a legal problem that needs to be resolved. - Pedro Beltrao
Lots of interesting lectures on the subject at http://lectures.gersteinlab.org/ - Roland Krause
Roland Krause
SS07: The future of scientific publishing
Robert Murphy: Mining images and caption from the scientific literature - Roland Krause
Images should be understandable without the text - Roland Krause
Examples: different colours, blank fields (information!), non-regular grids, graphs and images in the same grid - Roland Krause
(missing due to slow connection) - Roland Krause
Hand label, supported by machine learning, store in data base. Recognize scale bars, type of image, find the references between caption and figure. - Roland Krause
SLIF: Subcellular location image finder http://murphylab.web.cmu.edu/service... - Roland Krause
Interrogate the database by protein, location, cell type, resolution, GO terms, etc. - Roland Krause
Future: more automated system, improve practices for defining figure contents in papers. - Roland Krause
Mentions the Elsevier journal contest - Roland Krause
Future: Structured digital caption - hidden XML tags, includes coordinates, panel label, panel type, annotations, scale and length units - Roland Krause
Use of multilayer figures should be included with archival XML - Roland Krause
Very interesting work creating value, certainly underused - Roland Krause
The presentation is available too http://murphylab.web.cmu.edu/present... - Roland Krause
Michael Kuhn
PT47: BLASTing Small Molecules - Statistics and Extreme Statistics of Chemical Similarity Scores. Pierre Baldi
why is there no equivalent to BLAST for small molecules? is there a fundamental difference between small molecule similarity and sequence similarity? (PB think that seq. alignment even works for non-evolved seq., I tend to disagree --> site-directed mutagenesis!) - Michael Kuhn
reviews fingerprint method; standard approach: Tanimoto - Michael Kuhn
can come up with different chance models for fingerprint matches - Michael Kuhn
Given the diversity in tautomers, conformers, the variety in R-groups, etc, it's a different problem, IMO - Deepak Singh
[Huh, can't liveblog if the battery is empty] - Michael Kuhn
based on the ratio between two Gaussian distributions (one for the union and one for the intersection), can get p-values / Z-scores / e-values for a particular database - Michael Kuhn
evaluation: 55 known estrogen receptor binders x 100,000 random (presumably negative) compounds from ChemDB do leave-one-out tests on probability to get a false positive. Can fine-tune this to fingerprint size. - Michael Kuhn
[Editorial remark: I think this is very important work. However, I'm skeptical that it is a good idea to use biased chemical libraries for such analyses. For sequence BLAST, your standard is the genome which is under evolutionary pressure. Chemical libraries are not under such pressures, and therefore you can get "significant" hits in your database that are due to uneven sampling of the... more... - Michael Kuhn
Blast is acting on only a 1D string... There are many, *many* ways to define the similarity between molecules (see a review I wrote http://dx.doi.org/10..., but much more literature around). There is so much more complexity in small molecules than a amino acid sequence. Think if it as loop modeling for protein, but then without the luck of pinpointing the loop start and end to some homology modeled tertiary structure. - Egon Willighagen
Still, sounds an interesting read. Did Baldi mention a paper, maybe? - Egon Willighagen
yes, this is in the ISMB proceedings: http://www.ncbi.nlm.nih.gov/pubmed... - Michael Kuhn
Michael, thanx! - Egon Willighagen
Shirley Wu
Web 2.0 BoF session planning notes - http://docs.google.com/Doc...
The BoF is on Tuesday from 1-2:15pm in room 718B - Shirley Wu
Feel free to suggest or modify topics, volunteer to demo something, etc. I'm hoping this will be very interactive - Shirley Wu
Two years ago we already talked about Web 2.0 and actually gave a talk series at EMBL. Back then, we had RSS, delicious, citeUlike, but of course not twitter/friendfeed. I actually think that (as we've proven here) FriendFeed is a really effective way of forming an ad-hoc community, so I guess we should give another talk at EMBL. :) Anyway, we should give people this link as it's a Web 2.0 introduction for scientists: http://van.embl.de/cb... - Michael Kuhn
Hmm, I thought publishing it would make it editable by anyone but apparently not. Is there a way to do this? - Shirley Wu
This invitation link should work for anyone: http://docs.google.com/Doc... (I hope there are no spam-bots trawling for google docs links...) - Michael Kuhn
Cameron had an excellent blog on Friend Feed introduction for scientists (http://blog.openwetware.org/science...). I have a few other links related to the use and impact of FF on the blog (http://wildtype.wordpress.com/2008...) - Shannon McWeeney
If someone asks about Second Life we could show them Second Nature's homepage (http://www.nature.com/secondn...) where Nature hosts talks. A good example of a recent talk was the 23&me talk (http://scienceroll.com/2008...). I don't have a lot of experience in Second Life but I don't mind flying around if the connection is good enough. - Pedro Beltrao
One nice example of the use of second life in science is for archeology http://www.digitalhumanities.org/dh2007... - Pedro Beltrao
Session starting now - Pedro Beltrao
The google doc was updated as the session was going on with the gist of the discussion - Shirley Wu
Daniel Jurczak
“Pedro and Shirley” - http://friendfeed.com/e...
0.jpe
Struggling with the internet ;-) - Daniel Jurczak
Nice to see two shiny Macs on the desk - Michael Barton
Even though one of them was running XP ;) - Daniel Jurczak
Pedro Beltrao
HL40: A global analysis of genetic interactions in C.elegans (Matthew Weirauch)
Introduction to genetic interaction mapping in C.elegans and S. cerevisiae Charlie Boon Joal Bader, Roth , Trey Ideker etc - Pedro Beltrao
double RNAi perturbations to get genetic interactions in C.elegans. Quantified the interaction by progeny numbers - Pedro Beltrao
clustering of genetic interactions to determine functional cohesive groups (calls them congruency groups/networks) - Pedro Beltrao
He is following the structure of the JB paper linked above.Right .. this is a highlight session :) - Pedro Beltrao
Negative interactions tend to fall across pathways in worm - Pedro Beltrao
Are SGIs conserved across species C. elegans to S. cerevisiae. Not very conserved. Caveat mentioned that the interactions were targeted for metazoan pathways. Other possible caveats not mentioned (RNAi is knockdown not KO as in S. cerevisiae). Bridging between modules more conserved than individual interactions (hard to decouple between statistics and biological significance) - Pedro Beltrao
Shirley Wu
HL37: Atul Butte - Integration of genome-wide experiments to find genes associated with complex polygenic disorders: fighting back against the limitations of GWAS
Came in ~ 10 min late - Shirley Wu
Goal is to find candidate genes likely to be involved in complex disorders by integrating data from the many genome-wide functional studies already run. Two relevant questions - how to find _all_ gene variants relevant to a disease? How to find _novel_ genes/variants? - Shirley Wu
Using progeria as an example. Progeria = premature aging of children, very very rare, on average death occurs at age 15, 90% from artherosclerosis. But patients show loss of fat starting around age 2. Gene identified but obviously something very complex going on - Shirley Wu
RNAi study in worms - inactivating 112 genes increases fat content, inactivating another 305 decreases fat content - but worms don't even have fat cells. - Shirley Wu
compared 49 studies/data sets in a variety of organisms and modalities (mouse microarrays, proteomics, linkage studies, whole genome knockouts, worms, human) investigating fat cell or obesity to a knowledge gold standard (273 genes with variants known to be associated with obesity). Interested in quantifying the sensitivity of each individual study - Shirley Wu
Shows ROC curves for models based on individual studies evaluated against gold standard, then shows ROC for a simple model that just integrates all of the individual studies. Integrated model significantly beats the performance of all of the individual models. - Shirley Wu
Big take home point - all of these data were publicly available. Did not have to do any additional experimental work to produce a much better model for predicting disease genes. The real point is that combining experiments does better than single studies. This is the power of "integromics" - Shirley Wu
Ok, so we can predict known variants. What about discovering novel variants? - Shirley Wu
Talking about positive predictive value now. He likes PPV < 100% bc means the remaining % not in the gold standard positive set might be novel positives - Shirley Wu
Audience question: how does this method compare to traditional meta-analysis? Answer: still cutting edge, integrating different types of data across different species, so do need much more development. But naive methods so far are showing a ton of promise. - Shirley Wu
Audience question: how many genes do you think _are_ involved in obesity? Answer: if there's a long tail, well, there's a long tail! It's quite possible half the genome may be associated with obesity, with type 2 diabetes, aging, etc. These are very complex biological phenomena. - Shirley Wu
Shannon McWeeney
HL35: Pathway based diagnosis of cancer and metastasis
Talk by Han-Yu Chuang from Trey Ideker's group - Shannon McWeeney
issue again of breast cancer metastasis and poor ability to predict via expression (sample heterogeneity, poor correspondence among data sets). Key idea: overlaying relevant pathway information should improve results - Shannon McWeeney
Combined two data sets (van de Vijver et al NEJM and Wang et al Lancet) using overalpping gene set ~8000 genes. PPI data from public databases (Bind, reactome etc) - Shannon McWeeney
assessment of subnetworks (using greedy search for local maximum to id subnetworks) - considering as discriminant features - Shannon McWeeney
Search artifact therefore use permutation tests - global (to randomize relationship between expression and PPI). Local permutation test to handle topology bias. - Shannon McWeeney
Results are set of markers with functional relationships to each other - Shannon McWeeney
Reserved one data set for examination of classification accuracy. However, given choice of requirement of overlap at beginning, seems this could potentially over-inflate accuracy estimates - Shannon McWeeney
Discussion of other related work: use of module maps for targeted therapy - Wong et al. Cancer Research 2008 - Shannon McWeeney
Shannon McWeeney
HL33: Session Chair MIA - if anyone sees Yanay Ofran they may want to point him to the session. The speaker, Giorgio Favrin took it in stride and went forward with his talk.
nice - pn
session chair rushed in. looks like he was at the SS05 session with everyone else - Shannon McWeeney
Shirley Wu
TT21: Simbios - Fast molecular dynamics calculations on your laptop
molecular dynamics is computationally expensive on traditional chips. one way around this is to design your own chip. Another approach is to use graphics processing units (GPUs) - a la Folding@home on the Playstation3 - Shirley Wu
GPUs are advancing much faster than CPUs, currently you can do in a day on a GPU what would take a year or more on a CPU - Shirley Wu
Note: Simbios is a National Center for Biocomputing, so they make everything open and available for download and reuse. - Shirley Wu
OpenMM (open molecular modeling) project - library of methods for MD, code available for AMD and Nvidia GPUs. Isolates hardware-specific issues. Like mentioned before, currently being run on thousands of machines via Folding@home. Library currently links with Gromacs, and looking for collaborators for other MD packages. This project started by Pande group at Stanford - Shirley Wu
Showing a demo of CPU and GPU side by side doing MD calculations on one laptop (i think via virtual machine). GPU is chugging away, spitting out results for every 100 timesteps about 2x/second, it took the CPU a minute to spit out just one. - Shirley Wu
Audience question: if GPUs are so great, why aren't we doing everything on GPUs already? Answer: programming on the GPU is a pain right now. [Shirley's note: The industry is currently motivated by video game industry, hopefully this will change as other applications become profitable] - Shirley Wu
Now switching over to Simbody - open package for coarse-graining MD simulations as alternative way to speed up calculations. https://simtk.org/home/simbody - Shirley Wu
Uses internal coordinates to force certain groups into rigid bodies, reduce degrees of freedom by order of magnitude - Shirley Wu
Showing sample code for setting degrees of freedom, now showing a demo of 3 residues, first one with no reduction in df, second residue with Simbody defaults, and third residue where there are no df. - Shirley Wu
Now showing demo of a larger molecule with a helix set to rigid and other residues having certain degrees of freedom. This is good if you're not interested in what the helix is doing (assume it is rigid) but are interested in seeing how other parts of the molecule behave; it will speed up your calculations a lot and allow you to see much more. - Shirley Wu
Audience question: how does rigidifying parts of the molecule affect the dynamics of the non-rigid parts of the molecule? I didn't catch the answer, something about bond lengths - Shirley Wu
Pande group and Folding@home: http://folding.stanford.edu/Pande... - Shirley Wu
Simbody also allows you to create arbitrary molecules - Shirley Wu
Simbios (National center for physics based simulation of biological structures) website: http://simbios.stanford.edu/ - Shirley Wu
Disclaimer: my advisor is the PI on the Simbios grant, folks in my lab work on Simbody and SimTK. So I knew most of this stuff already but wanted to report on it to give it a plug for the rest of the ISMB audience :) - Shirley Wu
Talks like these make me weep. There I was a time I thought my SGI Octane was really fast. - Deepak Singh
I am lazy. Is Simbios an Internal Coordinate Dynamics program? - Deepak Singh
@Deepak, I think Simbios is the name of the center/organization. I'm definitely anticipating the release of OpenMM https://simtk.org/home/openmm to see how it compares to NAMD and others. - Adam Kraut
Right, Simbios is one of the NCBCs (National Center for Biomedical Computation) commissioned by NIH. SimBody is a library they've developed for coarse-graining MD simulations. Simbios also hosts SimTK (Simulation ToolKit), a server housing code and descriptions for dozens of research projects related to physics-based simulation of biological structures, everything from biomechanics of gait to molecular dynamics. - Shirley Wu
Oh that's very cool. Would definitely like to try that out - Deepak Singh
Michael Kuhn
Dopplr?
I know the conference is not quite over, but it'd be cool if we could connect in Dopplr to see when we can live-micro-blog the next conference :) http://www.dopplr.com/travell... (If you don't know Dopplr, it's a Web 2.0 service that lets you see if you happen to travel to the same town as someone you follow.) - Michael Kuhn
Adam Kraut
I summarized some keynotes from 3Dsig in a blog post; Thanks to everyone else for micro-blogging from the main event; it's almost like I'm still at the conference :) - http://www.bleedingedgebiotech.com/blog...
Shirley Wu
Would people interested in the BoF want to get together tonight? Are people planning to go to the CBW event?
CBW? I'd like to stay with poster but join you later possibly. Would rather meet around 8 and go for dinner. - Roland Krause
I would also like to browse around the posters a bit. Perhaps meet at 8 in the lobby of the Intercontinental and head off to the Elephant and Castle (?) for drinks and food and hopefully catch up with the CBW folks if they're still there. - Shirley Wu
That sounds great, count me in - Pedro Beltrao
Sounds great but I am going to see the Radiohead - Nosferatu mashup (http://www.blogto.com/photo_g...) at 9. I can't resist! - Shannon McWeeney
Ok, so plan is to meet in Intercontinental lobby at 8 - Shirley Wu
ok, 8pm - Roland Krause
Other ways to read this feed:Feed readerFacebook