 |
|
| FriendFeed is the easiest way to share online. Learn more » |
Adam Kraut › Likes
CASP8 Results | Foldit - http://fold.it/portal...
Paul Gardner, Adam Kraut, Wladimir Labeikovsky and 7 other people liked this
Summary of foldit results in the 2008 Protein Structure Prediction competition (CASP8): "foldit players are on par, but not better than protein folding experts at trying to solve the same problem with all tools available to them. It also appears that foldit outperformed all fully automated server submissions. Hopefully over time foldit can do even better, but being able to produce solutions of same quality as experts means that the top science research can now also be done outside of labs by game players, significantly speeding up the process of scientific advancement! "
- Michael Nielsen
PyRosetta - http://rosettadesigngroup.com/blog...
Michael Kuhn, Mickey Kosloff, Mitchell J Stanton-Cook and 2 other people liked this
Love how the textbook is published under Lulu. In the future, all textbooks will be published like this.
- Bosco Ho
The Way I Work: Jason Fried of 37Signals - http://www.inc.com/magazin...
Mickey Schafer, Adam Kraut, Konrad Förstner and 3 other people liked this
QUOTE "We rarely have meetings. I hate them. They're a huge waste of time, and they're costly. It's not one hour; it's 10, because you pulled 10 people away from their real work. Plus, they chop your day into small bits, so you have only 20 minutes of free time here or 45 minutes there. Creative people need unstructured time to get in the zone. You can't do that in 20 minutes."
- Michael Barton
I wouldn't complain if one day I was part of a company with the 37Signals mentality working in the field of bioinformatics.
- Michael Barton
Love this: "Creative people need unstructured time to get in the zone. You can't do that in 20 minutes." No kidding!
- Walter Jessen
Matt’s manifesto for a science data platform - http://mndoci.com/2009...
Adam Kraut, Khader Shameer, Pawel Szczesny and Neil Saunders liked this
Dynamic interactions of proteins in complex networks: a more structured view. - http://rosettadesigngroup.com/blog...
Mickey Kosloff, Adam Kraut, Pedro Beltrao and Pawel Szczesny liked this
PhyLIS - a free Linux distribution for phylogenetics - http://www.eve.ucdavis.edu/rcthoms...
Paulo Nuin, Khader Shameer, Anders Norgaard and 5 other people liked this
"PhyLIS is a user-friendly, free linux distribution for phylogenetics. Install it and you have an instant phylogenetics workstation. No downloading packages or messing with compilers, no configuring software, no worrying about small differences between systems that mess up your scripts. Simply install, sit down, and work. PhyLIS started during a period when I was acquiring several new computers for a large phyloinformatic project and grew tired of installing general purpose linux distributions, and then having to spend an hour or two reconfiguring everything and adding software on each new computer. I began developing scripts that would do some of this work for me, eventually they became overly-complicated, and I still had to carry around thumbdrives full of software or needlessly re-download everything. Eventually it just became more desirable to have an operating system that was specifically geared towards doing phylogenetics. PhyLIS is based on Ubuntu linux, a widely used...
more...
- Mike Chelen
from Bookmarklet
Computer-aided design of functional protein interactions. - http://www.ncbi.nlm.nih.gov/entrez...
Adam Kraut, Deepak Singh and Ruchira S. Datta liked this
Conformation Dependence of Backbone Geometry in Proteins - http://www.cell.com/structu...
Mike Chelen, Adam Kraut, Bosco Ho and Deepak Singh liked this
I am my own model organism: getting genotyped (Part I) - http://toddharris.net/blog...
Bill Hooker, Nakao M., Adam Kraut and 10 other people liked this
Cool - I used to read sequence on films as well :-) Genotyping will be on top of my wish list for Christmas. I need to decide on 23andme , DecodeMe or another supplier. I may go with DecodeMe just to support them....if they still exist...
- Nils Reinton
@Nils: I'll detail why I went (initially) with 23andMe in a follow up post. But cost, data availability, and data sharing options where the main reasons. Navigenics charges $199 after the first year for updated analyses. And like you, I just wasn't sure about Decode being around for the long haul...
- Todd Harris
Hmmm. Having myself genotyped vs having a lasik surgery to get rid of glasses. On a student budget, can't really afford either, but the latter would be more bang for the buck. Would be cool though. But not via the service my ex-employers back home in Iceland offer - financial situation pretty dire currently.
- 'Mummi' Thorisson
Think I will put genotyping on my xmas wishlist as well. @Nils, @Todd: I too have read sequence on film; we are really dating ourselves by admitting to this!
- Bill Hooker
It's going to be the nerdiest xmas ever! One of my most exciting moments at the bench was reading off a sequencing gel late in the night, finally coming across the first mutation in a gene I had been positionally cloning for years. It was at the limits of resolution but still clear enough to poke your eye out. Seems mundane now, huh? I still think the readouts from old school sequencing gels are amazing - such a tangible way of seeing the unseeable.
- Todd Harris
I wish I'd kept a few of my old gel films -- I'd like to frame one to hang on a wall at home. Nerd art!
- Bill Hooker
Me too! I had a boxes and boxes of autorads, lost somewhere along the way during the inevitable academic life-shuffle.
- Todd Harris
@Bill, @Todd I'll dig out some of my old ones and send them to you for christmas...if you're lucky you'll get some of those with smiley bands in the corners and edges of the gel :-)
- Nils Reinton
Large-scale data and text mining - http://www.slideshare.net/larsjuh...
 |
Mike Chelen, Adam Kraut, Peter Miller and 4 other people liked this
Bioinformatics, Genomes, EC2, and Hadoop - http://aws.typepad.com/aws...
Mike Chelen, Endre Sebestyen, Nakao M. and 2 other people liked this
Talks about JCVI AMI, and other Hadoop bioinformatics apps
- Deepak Singh
XSLT+MySQL=Append GeneOntology terms to TinySeq to GeneOntology - http://plindenbaum.blogspot.com/2009...
Mike Chelen, joergkurtwegner and Adam Kraut liked this
ICSB 2009 Grant Jensen How electron cryotomography will provide the ultrastructural information needed for spatially explicit systems analyses
Adam Kraut, Abhishek Tiwari and Neil Saunders liked this
lab webpage http://www.jensenlab.caltech.edu/
- Pedro Beltrao
systems biology should strive to take into account the spatial features of cells
- Pedro Beltrao
giving an example of how knowing the structural details changed the modelling of a metabolic network. carbon fixation reactions and localized in carboxysomes and they cluster inside the cell. this has implications for the models
- Pedro Beltrao
3d movie of the cell and its structures .... nice :)
- Pedro Beltrao
its not fair ... they can use movies ,,, its hard to beat that for interesting presentations
- Pedro Beltrao
summary ... bacterial cells are highly structured and this must be taken into account when modelling biological processes. showing several examples of structures inside bacterial cell.
- Pedro Beltrao
Upload to S3 from Snow Leopard - http://august.lilleaas.net/s3_uplo...
Adam Kraut, Wladimir Labeikovsky and kartik vaithyanathan liked this
The ability to execute ruby scripts out of the box from automator is new in Snow Leopard. So are the contextual services. Snow Leopard was released mere days ago, we can only begin to imagine the different ways people will make use of this new and powerful feature.
- Deepak Singh
Sumo: One-off EC2 Instance Lanching - http://adam.blog.heroku.com/past...
Jan Aerts and Adam Kraut liked this
Crunching Molecules and Numbers in R - http://www.slideshare.net/rguha...
 |
Abhishek Tiwari, Ola, Adam Kraut and 3 other people liked this
Ian Foster: What's faster--a supercomputer or EC2? - http://ianfoster.typepad.com/blog...
Tom Walsh, Adam Kraut, Euan and 2 other people liked this
That old (NCSA) benchmark is getting a lot of mileage! I appreciate the definition of 'fast' in this post - time from job submission to job completion. QBETS predictions are cool but I read it as the time it takes to complete your job on a (free) shared public resource. If I sit in my machine room at a terminal and no queue the big iron will win every time. However it's no secret that I...
more...
- Adam Kraut
Lots of data and the network - http://mndoci.com/blog...
Peter Miller, Jean-Claude Bradley, Adam Kraut and 2 other people liked this
Birds of a Feather session: Semantic Web-Linked Data, organized by Eric Neumann, in T5
Nakao M., Cass Johnston, Adam Kraut and 9 other people liked this
linked data is: a simple set of 4 guidelines for publishing RDF data on the Web (over HTTP), developed by Tim Berners-Lee in 2006
- Ruchira S. Datta
1. Use URIs as names for things (globally unique identity). 2. Use HTTP URIs (everyone has a web browser/client) 3. When someone looks up a URI, provide useful information...in the form of RDF data. 4. Include links to other URIs (foster discovery of additional information).
- Ruchira S. Datta
Context-independent identifiers (URIs) would make things so much more useful and interoperable - like Lego pieces.
- Ruchira S. Datta
Some want to get the semantics exactly right and use formal logic and OWL, but here we're emphasizing just the linkability of things.
- Ruchira S. Datta
A URI can only refer to one thing, but one thing can have several URIs, unfortunately.
- Ruchira S. Datta
several years ago, tried to bridge use LSIDs (life science ids): thing:something:something:identifier. But this can only be recognized by some particular software, not a web browser. Strong influence from W3C to use HTTP URIs, per the law of least power: do what requires the least technology. Even Mark Wilkerson who was touting LSIDs has come around to HTTP URIs.
- Ruchira S. Datta
A commenter says LSIDs still exist, it's just that they can extract them from HTTP URIs.
- Ruchira S. Datta
There are other proposals, e.g., shared names; Neumann prefers even less constraint than shared names.
- Ruchira S. Datta
So, now if you put in a URI you get something back. You should be able to get RDF back. UniProt does this: if you put .rdf on the end of the URI, you'll get the data back as RDF.
- Ruchira S. Datta
Now colleagues can just use the URIs in order to reuse the data; don't need to copy the data.
- Ruchira S. Datta
someone says the UniProt accession is an identifier, whereas the URI is a way to get at the thing through the web
- Ruchira S. Datta
identifiers can overlap, but HTTP URIs put things in unique namespaces
- Ruchira S. Datta
it needs to be stable: when you put this out, you're establishing a contract with the community that it's going to change
- Ruchira S. Datta
currently, the url, e.g., http://www.uniprot.org/uniprot... is also the URI. the second part is the identifier of the record and the part before the slash is the namespace. At http://purl.bioontology.org, we separate the namespace and the url. So going there we have a PO box that can eternally forward it.
- Ruchira S. Datta
Problem: this assumes http://purl.bioontology.org may go away. Thus the Banff Manifesto.
- Ruchira S. Datta
Reduce the likelihood of catastrophic failure by consolidating it into an institution, e.g., Stanford University, with longevity.
- Ruchira S. Datta
This just pushes the problem onto purl.bioontology.org.
- Ruchira S. Datta
The domain name can be transferred, so why do we need purl.bioontology.org?
- Ruchira S. Datta
Transferring a zillion domain names is a pain, transferring one domain name is easy. The institution commits to maintaining that domain.
- Ruchira S. Datta
It should be not just the institution, but the community--the community will continue to live on.
- Ruchira S. Datta
Knowledge should be monotonic: it grows and doesn't disappear. Even if a particular effort dries up, the URIs should still be valid so we can still see what was there.
- Ruchira S. Datta
The Linking Open Data Project: A community project started within the W3C Semantic Web Education & Outreach group in 2007
- Ruchira S. Datta
The LOD (Linking Of Data) "cloud", May 2007: many projects with various links between them, e.g., MusicBrainz, FOAF, DBpedia, etc.
- Ruchira S. Datta
By March 2008, had tripled
- Ruchira S. Datta
you can put any kind of data up and make it available to Sparkle queries
- Ruchira S. Datta
By September, WordNet and various other dbs had come in
- Ruchira S. Datta
March 2009: life sciences comes in, with Bio2RDF
- Ruchira S. Datta
now you can find the data that is in NCBI and UniProt in RDF format, but not the experimental data yet
- Ruchira S. Datta
to make this useful for interesting research, will need URIs, and to figure out what are the rules that are important for life sciences
- Ruchira S. Datta
when you publish using this data, how is your data that builds on top of it going to be able that's linked from it?
- Ruchira S. Datta
we don't really have this concept in life sciences yet, people don't know about it
- Ruchira S. Datta
suppose one looks for a concept in the LOD cloud, like "heart"; how do we know which thing to query? BioOntology, DBpedia, etc?
- Ruchira S. Datta
one can't do the Google on it yet
- Ruchira S. Datta
bioontology guy hates Google analogy; Google gives millions of hits, but we want the contextual query
- Ruchira S. Datta
i protest, have to have indexing before ranking
- Ruchira S. Datta
this doesn't solve the problem of redundancy: we want the facts about a protein, regardless of their source
- Ruchira S. Datta
bioontology guy says you don't need the index to answer the query, just to answer it fast
- Ruchira S. Datta
someone else says, need the indexing in order to do the clustering
- Ruchira S. Datta
she says you need semantic web overlays. we need hierarchical indexing environment in order to do this at scale
- Ruchira S. Datta
one needs to be able to query on an abstraction
- Ruchira S. Datta
bioontology guy: what's more important, query or browsing?
- Ruchira S. Datta
someone else: even browsing, if something is 3 links away, may not even go there
- Ruchira S. Datta
bio2rdf guy says: first we ask everyone simultaneously: do you know about this? then we ask what do you know about it? we have implemented Shared Names. But the URI just goes to the original record. Many people have said many things about the same entity.
- Ruchira S. Datta
nobody wants to have to read all the papers in MedLine. the punchline is the links: how does this protein relate to others. if we don't trust a link, *then* we want to drill down
- Ruchira S. Datta
if there are 5 million sources of "A is related to B", we don't want to read all of them, we just want to know that there are 5 million of them. we also want to know the kind of evidence, e.g., particular kind of experiment. Then the user can decide whether to trust it.
- Ruchira S. Datta
At this conference, enormous number of people mining data. We should be able to see their results as easily as the original sources.
- Ruchira S. Datta
Great coverage! Thanx! It's like being there... would have loved to sneak in on this BoF...
- Egon Willighagen
Egon: feel free to wander in...
- Ruchira S. Datta
we want just the local subnetwork, not the text. we want the facts
- Ruchira S. Datta
people want question answering
- Ruchira S. Datta
Bing bought Powerset for this purpose
- Ruchira S. Datta
someone says this turned out to be crap, e.g., "Psoriasis causes arms"
- Ruchira S. Datta
the question can be a small subgraph, not necessarily an English sentence
- Ruchira S. Datta
when we want information about a protein, there are only a limited number of kinds of things we can be interested in, so the software can guide the query context-sensitively
- Ruchira S. Datta
we need to distinguish the problem of document retrieval from query formulation
- Ruchira S. Datta
we shouldn't just think of scientists, but also other kinds of users
- Ruchira S. Datta
this will all be possible, but many people are currently just reinventing RDF over and over again
- Ruchira S. Datta
how many here are producers of RDF? roughly 8
- Ruchira S. Datta
put the things that we create in RDF, e.g., if you make the intersection of this fact with this paper, you are in charge of minting that URI
- Ruchira S. Datta
if everyone does this, then this facilitates cross-references and exchanges
- Ruchira S. Datta
http://bio2rdf.org querying life
- Ruchira S. Datta
Nophar Geifman has been working with Eytan Ruppin on finding cliques in GO around different diseases. A thing like that should have an URI, so other people can use it.
- Ruchira S. Datta
can we do micro-experiments so by ISMB next year, we can prove this concept
- Ruchira S. Datta
paradigm shift between hypothesis-driven query versus, the data throws the hypothesis at you
- Ruchira S. Datta
what's important is the use case: what is the question you can answer that would make them go "wow"?
- Ruchira S. Datta
David Hune is working with FreeBase and has developed Parallax, a facet browser
- Ruchira S. Datta
he also developed Exhibit
- Ruchira S. Datta
faceted browsing makes more sense to biologists
- Ruchira S. Datta
look at Google's Wonder Wheel
- Ruchira S. Datta
Jamie Gonagell (sp??!) at SciFoo camp designs games, first slide was World of Warcraft: if you harnessed the collective brainpower that youngsters spend on WoW every day, you could rewrite Wikipedia every day!
- Ruchira S. Datta
we need to figure out how to pull people in
- Ruchira S. Datta
I mentioned during the session, but forgot to link here (hard to talk and type at the same time!), Marti Hearst's new book _Search User Interfaces_ http://searchuserinterfaces.com
- Ruchira S. Datta
How to pull people in - that's the challenge to get this working. Could we get some seed money into this scientific effort and have it distributed with mechanisms similar to Google Adds? Yes if so we could have students and scientists putting efforts into this rather than some obscure webservers, blog etc with Google Adds. But how to generate the seed money? Government grants, donations or pay for usage?
- Bo Servenius
High performance computation and interactive display of molecular orbitals on GPUs and multi-core CPUs - http://www.ks.uiuc.edu/Publica...
Adam Kraut and Marcus D. Hanwell liked this
Thirty years of biomolecular simulation - http://mndoci.com/blog...
Peter Miller, Khader Shameer, Ola and 8 other people liked this
As computational power increases, I think we actually need more elaborate score functions (anybody said three-body terms ?!?) and not focus on soft potentials
- Nir London
Nir, I don't think our computers are powerful enough yet for more complex terms, but yes eventually. In the meantime, softcore potentials do allow us to address some interesting problems
- Deepak Singh
The most pressing need right now: reasonable models for chemical reactivity in approximate classical force fields.
- Dan Gezelter
@Nir, we certainly need better accuracy at the fine grain level so that simulations can reproduce enzyme mechanisms for example; but there's also the problem that some of the most biologically relevant space and time scales are not feasible even with petaflops of compute which is why coarse-grain methods are important too
- Adam Kraut
Speaking of petaflops; I think force-field development is going to get really interesting as some of the longest MD simulations are on the horizon (accelerators, massively parallel codes) which will undoubtedly show that current FF's break down or 'artifact' really bad when you run a system for that long. Tom Cheatham's (Amber/Utah) work has made a pretty clear case of this.
- Adam Kraut
What, Deepak? Computers aren't powerful enough? What about EC2 apps? ;-)
- Mr. Gunn
lol ... even Roadrunner is not a fraction of what we need
- Deepak Singh
I don't know a lot about atomistic simulations...but I'm going to learn a lot over the next couple years. Our DTRA grant lead PI is Susan Atlas and Steve Valone is also on the grant. They've been working on force fields using what they're calling charge transfer embedded atom model (CT-EAM). I think this is related to what Adam said above? You can see one of their papers here: http://arxiv.org/abs/0705.0857 Are any of you familiar with this CT-EAM work?
- Steve Koch
Interesting work. I am not familiar with it, but I am also a little away from the days I actively followed the field. The problem with proteins are that long range electrostatics are important in ways that polymers and other materials don't need to worry about.
- Deepak Singh
I'm mulling over a blog post about the fact that all MD packages fail badly in simulating the protein backbone. Very few people realize this and the reason I noticed it was because I spent grad school and the first 4 years of postdoc studying the protein backbone in crystal structures before moving to MD simulations 6 years ago.
- Bosco Ho
completely agree. That's my biggest issue with MD today. We need an upgrade to the physics. Back in the day I used to have to apply soft constraints, but that's a cop out
- Deepak Singh
from IM
Using RAID on EC2 EBS Volumes to Break the 1TB Barrier and Increase Performance - http://alestic.com/2009...
Deepak Singh and Adam Kraut liked this
Interested that you were feeling pinched by EBS performance. I need to read-up on the underlying architecture of those volumes. I would've guessed that performance concerns were unlikely, except under pretty substantial load. Have you run any performance benchmarks for your filesystems with/without RAID'ed EBS vols? I'm interested!
- Benson Miller
Benson: I've not personally experienced any problems with either size or performance (of EBS volumes) but once you start feeling limitations, is one way to alleviate them.
- Eric Hammond
Pittsburgh ranked tops in U.S. by The Economist - http://www.post-gazette.com/pg...
 |
Abhishek Tiwari and Adam Kraut liked this
I see Sydney in Austria did well too :-)
- Noel O'Boyle
Easy Map-Reduce With Hadoop Streaming - http://www.igvita.com/2009...
Adam Kraut, Todd Hoff and Abhishek Tiwari liked this
Automate EC2 Instance Setup with user-data Scripts - http://alestic.com/2009...
Deepak Singh, Abhishek Tiwari and Adam Kraut liked this
Next Point of Server Differentiation: Effiiciency at Very High Temprature - http://perspectives.mvdirona.com/2009...
Amund Tveit and Adam Kraut liked this
"Make the server suppliers compete for work done per joule at 95F approach temperatures and the server world will evolve quickly. It’s good for the environment and is perhaps the largest and easiest to obtain cost reduction on the horizon".
- Deepak Singh
Data center n00b question, Deepak, but is it possible for two systems to have completely different comparisons at two different temperatures? (e.g. system I is vastly better than system II in "standard" A/C-d environments, but II is much better than I at 95 F)
- Benjamin Tseng
I am no data center guru, but do have a clarification question. What are you comparing? Performance? Cost? Reliability? Note that sytems are getting very complex so at some level you can say "yes", since the design is likely to be temperature dependent
- Deepak Singh
Sorry, I guess I wasn't being clear on two levels: First, I was referring to the metric you seemed to be proposing (work per joule). Second, I was wondering if a work/joule test at 95F was good enough to measure server/data center differentiation. It sounds like your take is no -- as I could be tweaking my data center for something else, or a server-level innovation for max efficiency at 95F may conflict/have second order effects with data center "systems" innovations. Did I interpret you correctly there?
- Benjamin Tseng
Should have put that thing in quotes, since that's James' metric not mine. I believe what James was saying that there are significant cost advantages to running data centers at high temperatures given the cost of cooling, but to get there we need to do some design work and you need to get maximum output from your designs. A system designed for cooler temperatures will not work as well at higher temperatures and vice versa.
- Deepak Singh
High Energy Hadoop http://www.cloudera.com/blog...
Anil Thomas, Adam Kraut, imabonehead and Rajarshi Guha liked this
"how Hadoop is being used to process the results from High-Energy Physics experiments"
- Abhishek Tiwari
from Bookmarklet
Hadoop, Chunks and Multi-line Records - http://blog.rguha.net/?p=310
Pierre Lindenbaum, Amund Tveit, Adam Kraut and 2 other people liked this
Other ways to read this feed:
