 |
|
| FriendFeed is the easiest way to share online. Learn more » |
PSB 2009
Room for discussion and recording of events surrounding the Pacific Symposium on Biocomputing, Hawai'i, January 2009.
safe and easy way to buy apple ipad 2 and its accessories... http://gilanggumilar.com/
The need for a clear identity to aggregate contributions - openid, researcherid etc
Mr. Gunn liked this
Oleksii Kuchaiev - Learning the Structure of Protein-Protein Interaction Networks
Can someone correct me, but didn't someone just yesterday show a proof for why it's impossible for geometric graphs to capture the structure of PPI data?
- Paul J. Davis
I think they showed that random geometric graphs can't - i didn't catch if this one is random or not
- Shirley Wu
but if so, that could make for an interesting discussion session
- Shirley Wu
The proof I saw was that geometric graphs can't represent the bipartite graphs which supposedly is fairly common in PPI networks. My knowledge of the subject extends as far as yesterday's talks though, so I could be at a disadvantage.
- Paul J. Davis
Also, can someone describe to me what the general idea is with everyone trying to *generate* PPI networks from properties? Is this to try and determine false positives/negatives in other data that's less good or some such?
- Paul J. Davis
The "general idea" is the following: To do any statistical test you need three components: your data/hypothesis; statistical method and a valid null model (or random model). The focus of this talk was how to get better null models for PPI networks. If your null model is bad, any p-values (or other stats) you compute using bad model won't just mean anything. There are also other applications of models for PPI networks, other than using them in statistical tests
- Oleksii Kuchaiev
@Oleksii Ahah, that definitely makes alot more sense now.
- Paul J. Davis
Open science workshop recap - BMIR research colloquium - http://www.slideshare.net/shwu...
 |
from: shwu 25 minutes ago Recap of the PSB 2009 Open Science workshop, given at the Biomedical Informatics Research group weekly colloquium, 1/15/2009 Tags: psb-09psb09-openscience
That's it from me folks - I am off Volcano hunting and then on to LA and North Carolina!
Michael Nielsen liked this
Thanks for your efforts on the workshop. Have fun with your round the world trip.
- Paul J. Davis
Enjoy !! @Cam - Thinking ahead to Henry Gee's CISB'09 in Cromer, we have to work on liveblogging etc. With Mogulus, is there a cost to using it for live-streaming?? If not, I'll arrange to live-stream relevant sections of the event.
- Graham Steel
Mogulus is free for live streaming, that and Ustream.tv are probably the most popular ones now
- Mike Chelen
Thanks Mike. I've set up a Mogulus channel http://www.mogulus.com/steelgr... and am learning the ropes
- Graham Steel
This hasn't come up yet, but I wonder if there are any issues with microblogging talks in that people don't expect what they say off the cuff to go outside the walls of the conference necessarily - at least not like this. Also related, potential problems with accuracy of transcribing?
Cameron Neylon, Pedro Beltrao, Jim Hardy and Roland Krause liked this
I meant us microbloggers recording what the speakers are saying and the speakers not necessarily aware that their every word could be broadcast to the entire world
- Shirley Wu
Thanks Neil. Though I suppose I'm going to get a biased sample by asking through FF! The people who potentially might have issues are probably those who don't use these types of services
- Shirley Wu
But this is a matter of (some) concern, particular for smaller meetings. Note that the Cold Spring Harbor meetings have a press section that explicitly forbid coverage in the media to ensure that the presentations can be labeled as scientific communication and do not collide with embargo policies. https://meetings.cshl.edu/meeting...
- Roland Krause
Probably gonna write an entire blog post around this, but the core idea I'm coming up with is this: Its a game of trust.
- Paul J. Davis
My suspicion it that the big concern is being "misquoted" and having no why to validate or substantiate what was really said. Puts a lot of responsibility on the microbloggers shoulders to be accurate.
- Jim Hardy
We were explicitly asked not to take pictures of slides - also why I haven't broadcast video for the rest of the meeting. I guess I could have broadcast but not recorded
- Cameron Neylon
I think we might hear more about this soon when more people realize that there are people doing live coverage and/or blogging about the meetings. It feels like an extension of the cultural problems of the digital world as detailed by Lessig. It is increasingly easy to cover live events. It is obviously useful but not everyone is aware that this is happening. It also poses additional problems for cases where the events are profitable and live coverage might lower demand.
- Pedro Beltrao
I think that's an important point Pedro - awareness. I think Russ Altman mentioned in an intro that it was happening but I don't know that the audience necessarily would get the implications of that.
- Cameron Neylon
Perhaps until this becomes more mainstream it would be a good idea for those hoping to provide live coverage to get explicit permission to do so - just to avoid any kerfuffles that might turn people off from this kind of thing. It's too late for this meeting and I didn't think about it until now but in the future...
- Shirley Wu
But how would we get the permission ? Send an email to the organizers? They would probably refuse just to be on the safe side. Maybe once there are a few more conferences with associated online notes and maybe with corresponding conference reports it would be easier to point to these and ask for permission. Otherwise maybe people would not take it seriously.
- Pedro Beltrao
ISMB is already a great example - and it's a prestigious enough conference that other conferences in the field would take notice. I'm still ambivalent about the permission issue.. will think about it more
- Shirley Wu
I would argue that the conferences should have a clearly articulated policy. Cold Spring Harbour are very clear for instance (in the wrong direction IMO) but other conferences could hopefully adopt different but equally clear policies and communicate this to attendees
- Cameron Neylon
@Pedro, Live coverage actually increases demand by drawing attention to the event. The TV cameras at the Superbowl don't make people any less interested in attending ;)
- Mike Chelen
Unison: An integrated Platform for computation Biology Discovery by Reece Hart, Genentech
Open access, open source
- Ling-Fung Tang
unison-db.org:5432 for direct access
- Ling-Fung Tang
Problems: always need to incrementally build a database: including importing public data, etc
- Ling-Fung Tang
Sequence analysis != sequence mining
- Ling-Fung Tang
sequence analysis, given a sequence, figure out what features are there, computationally easy
- Ling-Fung Tang
given a list of features, what're the sequences that match, much more difficult
- Ling-Fung Tang
unison, data warehouse. stock all public database, non-reluctantly: structure, sequence, annotations
- Ling-Fung Tang
update automatically with make files
- Ling-Fung Tang
mining for ITIM the old: 1. collect sequence 2. prune redundant sequences 3. for each unique sequence, predict a number of domains write a program that filters predictions 4. summarize hits with external data 5. do it again when source data are updated.
- Ling-Fung Tang
mining for ITIMs the Unison way using some specific procedure language like SQL
- Ling-Fung Tang
showing examples how unison facilitates complex mining
- Ling-Fung Tang
showing examples of unison web tools
- Ling-Fung Tang
showing exmaples of using unison as a platform for diverse tools with collaboration with some others
- Ling-Fung Tang
why unison? testing hypothesis should not be painful, otherwise, ppl dont ask it
- Ling-Fung Tang
esp for biotech company, we want to encourage ppl to test out their hypothesis
- Ling-Fung Tang
Example: mine for sequences w/conserved feature, locate SNPs and domains on structure
- Ling-Fung Tang
Anyone want to share a ride to KOA from the hotel around 11:30am?
Friday or Saturday? I'm driving down around then on Saturday
- Paul J. Davis
Friday - today. I already reserved a shuttle
- Shirley Wu
Thanks for the coverage. PSB has a terrible reputation for being a cheap holiday excuse with empty auditoriums in many European bioinformatics labs. From the remarks in this room, it appears to be an interesting conference though worth attending, at least virtually.
Deepak Singh liked this
This is my first time to PSB so I can't speak with much authority, but depending on whether the sessions are relevant to your area of research (varies every year) it seems it can be quite worthwhile. You do get ample opportunity to converse with big name scientists, and the discussion sessions following the talks have the potential to be very stimulating. On another note, I certainly wouldn't call PSB "cheap"! ;)
- Shirley Wu
GECCO paper deadline extended to Jan. 28. Conf. In Montreal this year.
Our paper on ant colony optimization for genetic analysis was accepted for EvoBIO'09.
PSB09 Open Science Workshop - Report to conference - http://www.slideshare.net/Cameron...
 |
Ricardo Vidal and Jim Hardy liked this
from: CameronNeylon 43 minutes ago A summary presentation to be given to the conference on 8 Jan at 12:30 Hawaii time. Central themes that came out of the workshop Tags: open-sciencepsb09-opensciencepsb09
Open Science workshop highlights: Cameron Neylon
Ling-Fung Tang, Michael Nielsen and Deepak Singh liked this
Brief description of the goals of open science - making outputs of research more available to more people more quickly and effectively
- Cameron Neylon
I find it amusing that Cameron is live-blogging from the stage.
- Paul J. Davis
Now motivating the problem: makes research better, publicly funded research --> publicly accessible, etc
- Cameron Neylon
Yes, it's cameron 2.0
- Cameron Neylon
(aka shirley on cameron's computer)
- Cameron Neylon
Now briefing on what we discussed during the workshop. Tools, funding, policy. Phil Bourne's talk about making science communication more connected and conducive to education/research
- Cameron Neylon
Dave de Roure's talk on myExperiment, which is a tool for sharing workflows. Also talked about how it allows sharing of other digital objects, and the issues surrounding building an online community and a user-oriented tool
- Cameron Neylon
An important point that comes up is motivation. The story about creating a myexperiment bootstrap by paying contributors sounds like it was a successful method for creating that initial core community
- Paul J. Davis
Quo's talk about lessons from Wikipedia. Points out that Wikipedia works, but when we try to apply the same approach in science, it often fails. Because it depends on millions of users and a small percentage of contributors. When the user base shrinks to a few hundred, the number of contributors shrinks too - often to just a handful. If you build it, they won't come. the user base is important.
- Cameron Neylon
Nigam Shah's talk on bio-ontologies. A plug for the Bio-Ontologies SIG at ISMB 2009 in Stockholm - theme is wikis in science
- Cameron Neylon
Another important point is the fact that if we extrapoloate linearly from wikipedia's contributor levels, most biological wiki's would have a contributor level of less than one.
- Paul J. Davis
Some initial reflections on the first 4 talks: design of tools can't be divorced from social issues. A great tool isn't the be all and end all. Community building is just as important if not more, need active conversation between all parties
- Cameron Neylon
Heather's talk about measuring the adoption of open science. Interesting findings about barriers to sharing - many people don't share because they don't think anyone will use it.
- Cameron Neylon
One other way to increase the ratio of people participating would be to increase the incentives. Say for example that versions of parts of the wiki could be submitted for publication periodically.
- Pedro Beltrao
Major conclusion from the session: we need to measure these things. Does sharing and openness provide benefits? what is the return on investment? difficult to measure but without quantitation will be difficult to persuade and change how things are done
- Cameron Neylon
Mentions where to find all the material for the workshop. http://tinyurl.com/psb09-o.... Slides, recorded video, friendfeed
- Cameron Neylon
main themes from session: if you build it they won't come. build the service AND the community. standards and methods of citation are at core of good science. need to do better at citing services, infrastructure, assistance. specific tools required (persistent identity, good repository systems)
- Cameron Neylon
going forward: focus on specific actions with measurable outcomes, identify successes and failures, celebrate and learn. remember that change is slow and there is time to reflect, think, learn and plan
- Cameron Neylon
draws analogy to a target with scattering of shots - this is where we are right now but we can make it more focused
- Cameron Neylon
"Improving the research process is an area for (experimental) research that requires the same rigour, standards (and funding) as anything else that we do." a statement inspired by a comment made by Drew Endy during the workshop
- Cameron Neylon
No questions from audience - BBQ outside too distracting ;)
- Cameron Neylon
We are aiming to livecast the workshop report to the conference in 45 minutes time (12 noon Hawaii time)
Pedro Beltrao and Jean-Claude Bradley liked this
Having to make a choice between recording a screencast and livecasting by the looks of it
- Cameron Neylon
Very tough. I would opt for recording a screencast, Cameron
- Graham Steel
I'm figuring if we get audio and the slides are up I can always put together a slidecast later (I will also record the audio independently I think)
- Cameron Neylon
incidentally the reason for this is that Shirley's macbook doesn't recognise my webcam - the logitech webcams apparently only talk to very recent macs
- Cameron Neylon
p.s. how is the audio at the moment?
- Cameron Neylon
http://www.mogulus.com/cameron... the sound is clear
- Pedro Beltrao
excellent
- Graham Steel
you are currently getting a discussion session on multiscale modelling
- Cameron Neylon
it is hard to see the screen on the video - at least with a screencast you have a guarantee that will always be good
- Jean-Claude Bradley
In the spirit of live-ness, does anyone have a questions to assail Cameron and/or Shirley with?
- Paul J. Davis
that's fair - my slides are up at http://www.slideshare.net/Cameron...
- Cameron Neylon
We can also move the camera to get a better view
- Shirley Wu
How is the video/audio? (this is Shirley)
- Cameron Neylon
great
- Pedro Beltrao
Multiscale modelling of peptide fragments in the ribosome tunnel
Peptide sequences that stall translation in the ribosome, macolide antibiotics that do the same for which crystal structures exist with macrolide in the ribosomal exit tunnel
- Cameron Neylon
Want to do simulation to efficiently smaple conformational space, sequence, and chain lengths, using a distributed approach. Coarse graining alone is not good enough because the chemical details of binding are very important eg water
- Cameron Neylon
tiem scale of peptide conformational rearrangement is also long - therefore need a multiscale approach
- Cameron Neylon
Pointing and "structyre and dynamics of ribosome bound water" Lucent, Pande etc
- Cameron Neylon
uses a statistical process - then some maths which i missed due to network issues. Something about a correlating factor that can capture long distance coarse grain issues
- Cameron Neylon
essentially a combination of fragment MD (parallel processing) and then CG monte carlo simulations
- Cameron Neylon
now referring to Petrone PNA 105:16549 - building free energy maps of the ribsome tunnerl via exhaustive umbrella sampling approach.
- Cameron Neylon
want to move this into the coarse grain chain based model. Have dihedral plus sidechains, plus sterics
- Cameron Neylon
showing a movie of a corase grain sampling exploring a free energy map derived from atomic resolution MD
- Cameron Neylon
still early days but hoping to publish in next few months - have shown some sequence specificity in terms of preferred binding modes to ribosomal tunner
- Cameron Neylon
Also using CG models to seed MD
- Cameron Neylon
Summary - can overcome sampling barriers, is ideal for distributions, samples across different chain lengths and sequences
- Cameron Neylon
Heading to the pool bar for beers if anyone's interested.
Keynote: Greg Hampikian: "DNA Don't Lie": How Bioinformatics freed some of my best friends, and sent the guilty to prison
Roland Krause liked this
The speaker is gifted at being able to speak to many different audiences - e.g. getting both 5 yr olds and cops to understand something like DNA and genetics
- Shirley Wu
Wrote a book called "Exit to Freedom" about man named Calvin C Johnson Jr who was arrested for two instances of rape based on witness identification in photos. He'd been arrested for home burglary a while ago but otherwise clean. Would have been convicted for 7 years but turned out the attacker was uncircumcised and Calvin was circumcised.
- Shirley Wu
Works in area of criminology
- Shirley Wu
Point is that gut feelings are hard to shake without hard evidence.
- Shirley Wu
Showing Home Infidelity PSA Semen Detection Test Kits which they use in court apparently. For when you think your wife is cheating on you
- Shirley Wu
Uses prostate-specific antigens to detect presence of male fluids in women's panties (#1 tested item)
- Shirley Wu
(Quite risque material, appropriate for after dinner...)
- Shirley Wu
"and what you want to know is is your Y chromosome the chromosome that is in her panties?"
- Shirley Wu
aaannnd moving on to next example: FamilyTreeDNA website
- Shirley Wu
And now back to Calvin C Johnson's case. the "Innocence Project" - something to do with proving people's innocence who were wrongly convicted using scientific evidence
- Shirley Wu
Apparently Greg Hampikian was mistaken for Barry Scheck in an interview (Barry Scheck (sp?) founded the Innocence Project)
- Shirley Wu
Describing examples of people convicted or acquitted because of DNA evidence. E.g. cop comes home, wife is dead, she'd had sex, turns out cop was having affair so he becomes primary suspect, he's convicted and goes to prison ("and you can imagine what it's like for a cop in prison.."), a while later the semen sample comes back from the lab and turns out it belonged to a serial rapist. He is released but they still won't reinstate him as a police officer
- Shirley Wu
Showing plot of # of death penalty exonerations by state. FL = 21, IL = 18. Since 1973, 122 people in 25 states released from death row with evidence of their innocence
- Shirley Wu
Avg # yrs btwn sentencing and exoneration = 9.2 yrs. One guy, Charles Fain, on death row for 18 yrs for rape and murder of 9 yr old girl.
- Shirley Wu
# of cases in which DNA played substantial role in establishing innocence: 14/122
- Shirley Wu
And now back to Calvin C Johnson case for real
- Shirley Wu
Accused of 2 separate rapes in two different counties. Situations of each case the same so they knew same person did it. Convicted of both. Insisted on second trial but no one would represent him.
- Shirley Wu
But managed to get second trial and that time he was found innocent. Differences? Racial makeup of jury and attorneys
- Shirley Wu
Now showing evidence presented. prosecution - both victims picked Johnson out of photo line up. Beard description - "no beard but could be stubble" - old photo of Johnson used where he didn't have beard though at the time he had full beard
- Shirley Wu
Pubic hair also found, prosecution claimed came from Johnson but defense said lab found that though hair from negro, not from Johnson
- Shirley Wu
Blood type O+ - matches Johnson, but also matches 30% of black population.
- Shirley Wu
In live line up, the victim chooses someone other than Johnson. But at trial, she claims she knew who the attacker was but chose someone else
- Shirley Wu
More examples of wrongful convictions, often with many years of the sentence served before exoneration
- Shirley Wu
Lauding those who preserve evidence for 10, 20, + years. Amazed that after all those years, you can still use the samples to do DNA tests and to identify people
- Shirley Wu
Surprising conclusion from Innocence Project: fraudulent science played role in 34% of prosecutions
- Shirley Wu
63% due to errors in forensic testing (some of these errors are that the test was never even done but testified as having been done)
- Shirley Wu
Now describing some of the terrible ways in which line ups are conducted, leading to faulty ID
- Shirley Wu
In general, 25% of the time, the primary suspect is cleared of wrongdoing by DNA evidence
- Shirley Wu
Differential extraction to get sperm cells out of heterogeneous mixture - possible because sperm cells have lots of heavy disulfide bridges
- Shirley Wu
Running through some basic biology... a little unnecessary for this crowd
- Shirley Wu
Going through paternity cases that use DNA tests
- Shirley Wu
Talking about Short Tandem Repeat regions, test at 13-16 loci
- Shirley Wu
Apparently N=203 in caucasian database for STRs at those loci (seems surprisingly low to this audience)
- Shirley Wu
some problems with DNA - easy to transfer and contaminate. Mixtures not interpreted consistently. Statistical abuse
- Shirley Wu
Another rape, another murder
- Shirley Wu
Examples of DNA contamination problems incriminating wrong person
- Shirley Wu
Pretty much a talk full of sordid stories :P
- Shirley Wu
Oop, now talking about privacy ramifications of complex SNP analysis, modifications to GWAS data access
- Shirley Wu
Talking about the question of whether you can detect a person's ethnicity from their SNP profile. pretty controversial (but he doesn't discuss it)
- Shirley Wu
Something about using dog mitochrondrial DNA to implicate someone. But this led to idea to identify non-scooping dog poop violators using poop DNA profile
- Shirley Wu
Showing clip from Colbert Report: interview with Jerry Miller, the 200th person released from prison in Innocence Project
- Shirley Wu
Stephen Colbert gives Miller a card that says "I'm so sorry, - Sincerely, Society". Miller says, "Thank you, Society"
- Shirley Wu
And that's the end. Time for audience questions
- Shirley Wu
Audience questions: how did you choose the 9 out of ~ 3000 cases for the Innocence project?
- Shirley Wu
Answer: criteria about in-state, whether there was DNA, claim of innocence bc of faulty ID, etc
- Shirley Wu
Question: how to deal with increasing false positives as database size increases? Answer: (didn't catch a good answer) conversely many many cold hits that are not followed up bc case was too long ago, no more witnesses, etc
- Shirley Wu
Question: will forensics embrace high-throughput sequencing? Answer: need to sell the technology to them, the criminologists aren't reading all the latest papers to find the latest technology. No current push towards high-throughput sequencing
- Shirley Wu
Question: mentions chimeric people - people who have different DNA in different tissues. Answer: # of these people not clear but has ramifications not just for forensics but for organ transplants etc. Well-known case was woman trying to prove maternity. Chimeras higher rate in in vitro fertilization, two embryos take
- Shirley Wu
Panel session on molecular bioinformatics for disease
Khader Shameer, Tom Tullius, Chris Miller and 10 other people liked this
Califano: computational scientists should spend more time learning about the biology to some extent and vice versa - will produce higher impact research
- Shirley Wu
Trey Iddeker: In-house integration of computation + bio vs. straight collaboration where the two are pretty separated? Depends.
- Shirley Wu
Russ Altman: agree for the most part but behoove us not to overlook the algorithm for the biology or vice versa. If there is a good biological question you should see it through, and same for the algorithm. Wary of pressure being placed on young researchers now to excel in two areas. Rather expect them to be good in one area and know how to collaborate. Will lead to faster career and research advancement
- Shirley Wu
Larry Hunter: quoting Marcy McClure - "Bioinformatics is getting new science out of other people's data". Biology, success is all about creating new data. So it's hard to convince others of validity of Marcy's view.
- Shirley Wu
Learn how to think like a biologist. Doesn't mean you need a lab. Just get some intuition. Learn a very small slice of biology really well and develop methods to study that biology really well. Then you can move on.
- Shirley Wu
Ted Shortliffe: I'm actually going to disagree with a lot of what's been said - Sorry I'm an interloper, was actually at another meeting prior to now. Russ: "Call security!"
- Shirley Wu
(By the way, I didn't catch it but I believe the question they are asking is "advice to those entering the field" or something of that nature)_
- Shirley Wu
being good in the two areas (lab+maths/computer) is not really easy. These are very different mindsets and I know of very few (actually maybe just one) PIs that actually have working experience on both.
- Pedro Beltrao
Ted Shortliffe: Need to have a love and appreciation for the informatics. These methods could have really broad applicability (not just biology or medicine) but they get overshadowed by the flashier biology
- Shirley Wu
Califano: Could be really valuable to treat biological systems as informatics systems to glean new insights. COmputational scientists think about experiments differently, can be useful to think about it their way
- Shirley Wu
Idekker - wants to see bioinformaticians get more involved in the planning of experiments or studies. Analogy to EMT called to scene of accident trying to piece together what happened
- Shirley Wu
Audience member: collaboration is easy if you're a big name, but much harder for most of us. Even if you have a great tool, really hard to find an experimental lab willing to test it. Claim: need a different type of experimental science or we need to make our own
- Shirley Wu
Russ Altman: depends on choice of problem you're trying to solve, and planning in advance. Might be unfair expectations/pressure for experimentalists to get out of the blue request. Choice of problem - what if you have great algorithm but no existing dataset to do "rediscovery" type validation? It helps to have a plan at the outset for validation for it to be successful.
- Shirley Wu
califano: you could turn it around. Experimentalists come to informaticians all the time and say "i have this microarray dataset, can you help me analyse it?" So we need to be diplomatic, learn each others' language, invest time and effort in collaborations up front
- Shirley Wu
One problem with that is that in many cases , top-down research leads to biological predictions that can be related to many different fields so it is not easy to plane in advance for collaborations.
- Pedro Beltrao
Shortliffe: experience of bioinformatics students going to work in wet labs that have never had a bioinformatics student before. Mindset of wet lab at the beginning is "great, I have an IT techy person to help out now". But the good students transform the lab's thinking, show them what bioinformatics research can really do. Point is that students, not full professors or big names, CAN have a big impact in people's attitudes
- Shirley Wu
Idekker: observes that to get faculty positions in good bio programs you need to have familiarity with computational tools. So attitudes are definitely changing
- Shirley Wu
Question: suppose collaboration is successful. Then what? What do you do with the data? Where do you publish? Is there a bias against the methods/bioinformatics approaches in top journals (e.g. cell/nature/science)?
- Shirley Wu
Idekker: yes there are examples of papers where the bioinformatics methods or analyses or even results are in the supplementary table 7.2.a. But also examples of papers where the bioinformatics really made the paper. It's our choice.
- Shirley Wu
Russ Altman: "you say you have a great result." Is it a great biological result? or a great informatics result? Use this to determine where to publish. Chip Lawrence's Gibbs sampling paper in Science as example of how the informatics result was THE important.
- Shirley Wu
Hunter: could also try publishing two separate papers to highlight the two different aspects of the research. Doesn't have to overlap that much or count as redundant. Both are important.
- Shirley Wu
Audience followup: So then you hav Nature papers and Nature Methods papers. What about when someone looks at your CV and they say "we wish you had more Natures and less Nature Methods".
- Shirley Wu
Shortliffe: well, who are you trying to impress? informatics people or biochemistry people? Depends on the field too - you could publish a hugely important method which won't be in a big impact factor journal but a ton of people will read it and reference it because it's really useful - so you've had a huge impact on the field if the field is small
- Shirley Wu
(hmm, more tension between the establishment (IF) and more heuristic types of credit and evaluation)
- Shirley Wu
Russ Altman: even if you have a million Nature papers as middle author, you will be asked what was your contribution, and that's what matters. Fewer Nature papers but more important contributions will go farther.
- Shirley Wu
Steve Brenner asks panel: At biology conferences, the panels revolve around how to the protocols work, the results, analyses, etc. At bioinformatics conferences the panels (and especially here at PSB) they talk about "how do we do our science better"
- Shirley Wu
Russ Altman: "can you say some more about what the biology folks talk about?" Steve: "you know did you put reagent X in solution Y, etc, things that are -". Altman: "Trivial"
- Shirley Wu
"people who are faint of heart should not enter this field because the boundaries are unclear, constantly being defined, etc" If you want more well-defined boundaries then go into a more traditional field
- Shirley Wu
Hunter: show of hands - who is in an actual "bioinformatics or medical informatics department"? (small smattering of hands) "And most of those were probably founded by Ted (Shortliffe). So you're not being judged by your peers." which is part of the problem.
- Shirley Wu
Shortliffe: telling a story about Russ Altman (Shortliffe was faculty when Russ was an applicant for the MSTP program at Stanford). Russ was interviewed by one of the current students on the applicant selection committee.
- Shirley Wu
Student summarized Russ's case by saying "when he was at harvard he was in a great lab, but when you talk to him today, he has this interest in computers which kind of worries me but I believe in his heart when he gets here he will do great science." Reflecting that back then and even now, fighting to defend that what we're doing is science, mixing computer science with biology.
- Shirley Wu
Russ: "as you can see, I tricked them and it worked".
- Shirley Wu
Steve Brenner: rhetorical question. "if what we're facing is endemic to the nature of our discipline, are there models of other disciplines that have come before us that we can use as a template rather than having to pave our own way?"
- Shirley Wu
Califano: Corollary to genetics? Almost all the other disciplines if you think about it have been revolutionized by computation. In the future, expect to have all experiments in silico first.
- Shirley Wu
Hunter: likes to use biostatistics as model. Methods development, ethos of service and collaboration, considered successful and so it's possible. (ed. Not sure if i like that example)
- Shirley Wu
Ideker: would welcome discussion about methods, results, analyses, rather than ethereal meta-discussion. Suggests Steve Brenner plant questions in audience at bioinformatics conferences in that vein to redirect disucssions when they drift towards the meta-discussion.
- Shirley Wu
Question about funding. How to sell our research effectively, etc. Driven by the funding to pursue specific problems when our methods could be applied to many different things
- Shirley Wu
Hunter: huge investment in health information technology infrastructure coming up. HIPAA is a lesson for what can happen - collateral damage to research enterprise. What aspects of health technology infrastructure do we need?
- Shirley Wu
Sean Mooney: lots of investment in information cyber-infrastructure too. CTSA awards, etc. Comments that informatics is not getting any easier - how hard is informatics going to get as these systems come online, as storing and recording data get more complicated?
- Shirley Wu
Russ: agree, it is getting harder. Example with next-gen sequencing. We thought we had it all under control, we have BLAST, etc. but then next gen sequencing came along and all hell broke loose and everyone is scrambling
- Shirley Wu
Battery running out about now, so will go quiet but I will try to take paper notes and transcribe
- Shirley Wu
Audience: informatics types like things well-defined. this works ok with molecular functions, but not with diseases - hard to define phenotypes, molecular basis, etc. Where is the unification of ontology/definition w/disease so we can bring in effective informatics for clinical applications?
- Shirley Wu
10 second silence
- Shirley Wu
Shortliffe: and I'm stuck holding the mike...
- Shirley Wu
Audience member: current disease categories need not be fixed. Time for us to define new disease models that work better for us
- Shirley Wu
Russ: points out work by Atul Butte on defining new nosology (classification of disease) using molecular data
- Shirley Wu
Hunter: we're contributing to new definitions as our field uncovers new things about diseases like cancer, etc
- Shirley Wu
And that's a wrap. Time for dinner
- Shirley Wu
The unification IMO is statistics and harder math.
- Paul J. Davis
Thanks for the coverage Shirley it was an interesting session
- Pedro Beltrao
Thanks Shirley, very good of you to get this down for us
- Rob Syme
Oznur Tastan: Prediction of interactions between HIV-1 and human proteins using information integration
Pedro Beltrao liked this
current HIV therapies not optimal bc not accessible to all people, cannot eradicate from body, and drug resistance.
- Shirley Wu
Target the biology of HIV, which requires use of host cell machinery - need to predict novel direct physical interactions between HIV-1 and human proteins.
- Shirley Wu
Dyer et al Bioinformatics 2007, Davis et al Protein Science, Konig et al Cell 2008 previous work. But no study doing global analysis
- Shirley Wu
Supervised learning. Dah, that slide went by too fast. HIV-1 and human protein pairs represented with set of features
- Shirley Wu
Use decision trees, random forest classifier with majority vote out of possible trees
- Shirley Wu
Training and test data taken from keyword search of proteins and relationship terms. Differentiated between direct relationships ("binds to", "acetylates" etc) and indirect ("regulates", "affects")
- Shirley Wu
35 features including expression data, PPI data, sequence motifs, etc
- Shirley Wu
motif example: does human protein contain ligand domain or belong to ligand protein class? does HIV protein contain binding signature?
- Shirley Wu
Also use similarity of HIV-1 protein to human protein's interaction partner - sequence, cellular localization, post-translational modifications, etc
- Shirley Wu
evaluation: AUC 0.92 and Mean Avg Precision 0.23.
- Shirley Wu
Results recapitulate observation that pathogens tend to interact with host proteins with high degrees of connectivity to other proteins
- Shirley Wu
Resulting predictions: Tat interacts with Pin-1?
- Shirley Wu
I'd be a bit surprised if that were true, since a quick look at Tat sequences in GenBank indicates that it contains only one consensus Pin-1 recognition motif (ser/thr-pro), and the thr at that position is not well conserved (it's not, for instance, present in the refseq).
- Bill Hooker
Also, Tat is one of those miserable proteins (small, charged, ?globular) that will *interact* with pretty much anything. Whether the interaction means anything is another matter. The NCBI Gene entry for HIV-1 Tat lists well over a thousand protein-Tat interactions...
- Bill Hooker
I didn't really catch it hence the question mark. It was definitely Tat interacts with something, which doesn't really help
- Shirley Wu
The issue about bioinformatics struggling to identify itself as a discipline came up in discussion at PSB today.
Been working on and off all day on how to get ReliefF scores out of RWeka. I can get the ranked attributes but not the scores???
I think this relates to my previous tweets about U.S. medical schools being run like businesses. You get no respect unless you PI grants.
I firmly believe students need to be trained in both areas to be effective bioinformatics scientists.
Usual discussion at PSB on the interface between biology and informatics. How to train students? How to collaborate? etc.
NYT article on R: http://www.nytimes.com/2009...
next talk on FastChi: an efficient algorithm for detecting gene-gene interactions by Dr. Wei Wang of UNC
@CameronNeylon Thanks for trying the Twitter->FriendFeed linkup!
FastChi developed for mouse genotype-phenotype studies. X^2 for relationship between binary coded SNPs and binary phenotypes
Other ways to read this feed: