Bosco Ho › Comments

Bosco, this is a superb idea. Along with starting up a new journal/software hybrid, it will be great if existing journals insist users to submit source code, executable or VM of a bioinformatics software / database / server to a centralized repository like 'biohub.org'. - Khader Shameer

This is a good idea. - Michael Barton

While not linked to an actual repository (but rather, provides a snapshot of the s/w and data for the article), Journal of Statistical Software, does pretty much this - Rajarshi Guha

I would take this further and the article text remains in the revision repo. The reviewers are sent to the article, not the other way around and it can be forked in just the same way the software can - Frank from iPhone

@Frank, this makes sense, since otherwise the paper would be static and refer to old versions. But then this assumes that as the s/w is updated, so is the paper - Rajarshi Guha

@Rajarshi not neccessarily the paper should state which version/revision it refers to. It does not have to keep up with the sw. That is what documentation is for :) - Frank from iPhone

me likey too - Deepak Singh

The more I think about it, the more I think some big-wig bioinformaticians should do a deal with Google Code to edit a journal. That might even align with Google Scholar. - Bosco Ho

@Frank, in that case, why bother with a VCS? Why not just put a tarball with the source code for the version that goes with the paper? - Rajarshi Guha

Great idea, but I can't see it working for data sets. Yes data sets evolve and should track provenance somehow, but having been in and around standards groups for some time now, this is an impossible task for a publishing group to take care of, especially considering the nature of big-data bioinformatics. Plus if goes against best practices for software source control (use factories, don't store your database...) - delagoya

There are some interesting and non-trivial questions around this kind of idea as to what peer review should look like. Should such a journal provide virtualisation environments so that the code can be run? Example data should be a requirement presumably? Are peer reviewers expected to evaluate code "quality". Anyone thoughts on this would be extremely useful...and help guide a project like this into reality. - Cameron Neylon

My answers to Cameron's points: (1) no, (2) yes, sample data would probably be used to run tests which should pass, (3) quality is somewhat subjective - minimum requirement should be that code runs and generates output as expected - but reviewers could certainly suggest code improvement where appropriate. - Neil Saunders

So if the answer to 1) is no, does that mean that you can't necessarily expect referees to actually run the code? Or compile it? Or just that you pick referees appropriately? Or conversely that "refereeing" becomes a process of building up enough positive comments or karma points in the repository...? It seems to me that you want to bring the best of versioning systems and best practice... more... - Cameron Neylon

Referees should certainly be able to run code - I'm just not sure that virtualisation through the web interface is the way to do it. Seems like an additional layer of complexity that might get in the way of making this idea work. - Neil Saunders

@Cameron & Neil: If it could be figured out how to to handle the virtualization (or having remote access to machines), I think that'd be a highly valuable addition to peer review. Easy for me to say (not knowing how to implement it), but I think it's a great goal to strive for. It doesn't seem too crazy to have the journal have a bunch of machines on hand so the authors can remotely upload / install code and referees could then remotely log in to look at and try out code. - Steve Koch

I can't figure out where to jump into this thread. Personally, I think we just need a place to publish locations, i.e. the code is here, data is there and this is the version we used, etc. That must be maintained and being able to maintain that should become part of the funding process. Since funding agencies are the ones who are funding this research they need to include the ability to... more... - Deepak Singh

My feeling is that being able to run the programs somewhere on a server without downloading them is important - but that is very much a user's perspective. I often look at useful things that are made available and just have no clue how to actually make them work. A good range of downloadable executables would probably do the job for me though. Additional question: what are the standards for web services? - Cameron Neylon

Which is why VM's and cloud services are such a big deal for demo's and provenance now. You can package up a VM with the exact stack that you want and make it available, either as a service or a VM you can launch yourself. It's too easy not to do it - Deepak Singh

@Deepak : Cloud + VM is an an interesting combination, but should have an accessible pricing that is affordable to a larger research community - Khader Shameer

I think there should be strict guidelines while reviewing bioinformatics software / database / servers to test the resource. I had a recent experience : a reviewer wrote extensive list of points to reject a server that we developed with out trying what exactly it is doing or to know how does it differs from other existing resources. I strongly support the hybrid journal model, also it... more... - Khader Shameer

Let's talk specifics. VM images are great, but you are tying your release to a particular release of a particular platform. A better approach is to start from a base OS (like a linus distro ISO) and have a set of build instructions for system set up and application building. My favorite of the moment would be Chef. - delagoya

Second, academics love to solve a problem with a novel algorithm and then move on. In fact it is in their best interest to move on after milking a project for all it's worth, publication wise. Maintenance, or even robust testing (couch... Tophat ... cough ... Bowtie .. cough ) is not even on the radar. Frankly I am not so sure it should be. Maintenance requirements may slow the pace of... more... - delagoya

@delagoya, good point. If I have made significant improvements, why update the old paper? better to try for a new paper! - Rajarshi Guha

delagoya, chef's fine too. Find a common medium/mechanism that works for the community. The resources are certainly there. It's a matter of trying things out. As someone I know says, start simple, and iterate - Deepak Singh

Khader, that's where the funding agencies come in. They need to provide mechanisms for sustainable funding here. - Deepak Singh

The nice thing about a hybrid journal is that it might be possible to have new dois/database entries for "significant" updates. Not perhaps just place holding papers as is the case sometimes in the NAR database issue but when something has changed significantly you can get a new paper without needing a new algorithm or service. I like the idea of funding to support "orphan" code and services as well. Make it worth money and people will do it. - Cameron Neylon

Delagoya - as a naive user I disagree. I really don't want to have to build, I want to use in the lowest stress way possible and a hosted VM seems like a good way to enable that - as well as allow for longer term preservation. We may not be able to run linux on future hardware but will probably be able to handle VMs for longer (actually having written that I'm not sure its true - would be interested in more expert perspectives) - Cameron Neylon

I almost missed this discussion. I really like the idea but I wonder how discovery type projects fit in. I mostly use code to look for trends. If anything I might make some predictor to enhance existing data. For these reasons most of what I do is one off scripts around perl and R. Maybe this sort of project does not belong in a bioinformatics journal at all. - Pedro Beltrao

Pedro, great question. Personally, if we included all glue code, small scripts, etc this would be unsustainable and defeat the purpose of peer review as well - Deepak Singh

@Pedro, I don't see a journal/software hybrid as replacing all bioinformatics journals. I think there's a place for journals that discuss pure algorithms and ideas. These would do exploratory type programming. Normal journals service these papers quite well. For me, a hybrid model targets specifically those papers that describe a program that is meant to be used by other people. In that... more... - Bosco Ho

Bosco, you're thinking along the lines of a communications journal aren't you. And then people can go to work on the code if it is on github or something - Deepak Singh

@Deepak. Yep. The disconnect I see is that pragmatically, it's the open-source project that counts. The article in the bioinformatics journal is so that we can get a place-holder to collect citations that contribute to our academic CV. The journal/software hybrid provides the most efficient way to this goal. - Bosco Ho

it is excellent. I'm not so sure about that part about chicken though. - Richard Akerman

What a great find! - Chris Lasher

Certainly seems to refute earlier work by C. Sanders. - Matthew Todd

Chicken. Chicken chicken chicken. - Bosco Ho

Rhubarb Rhubarb - http://en.wikipedia.org/wiki... - Andrew Lang

I've read about 40 books on the iPhone, both on the iPhone Kindle, on Stanza and Eucalyptus. I have Project Gutenberg available on 3G all the time, as well as the Amazon web store. With the backlight light, I can read literally everywhere. Font size is fine, and I can flip open a book with one hand, even waiting in line at the supermarket. - Bosco Ho

the lack of books and twice the price is what kills kindle in australia for me at the moment,,, - Terry O'Fee

Agreed. I think the iphone is being touted as a competitor to buy time until Apple can develop a real one .. a tablet perhaps. - Ahsan Ali aka. Slick

Those never having used a Kindle or a similar device probably underestimate the joy of reading a page rendered on flicker free e-ink. - Thomas Amberg

that's not the issue. i'd like to get one myself. it just seems every country apart from the US seems to get screwed over when it comes to ebooks in general.. - Terry O'Fee

Ahh, dates back at least to BOSC 2002 from a talk by Jason Stajich http://open-bio.org/bosc200.... I couldn't find it on bioperl.org which is why I thought I was remembering it wrong. But perhaps this has been retired as "official" Bioperl dogma? - Andrew Su

Ah there was formal dogma? I was almost done with Perl by 2002 :) - Deepak Singh

You know, I too was pretty much done learning perl in 2002, but the crazy thing is that my 2002-perl knowledge still allows me to do 90% of what I want to do. Anyway, hopefully Jason will chime in here with the official line... - Andrew Su

It hasn't changed much since 2002. I am pretty sure Jason has the official line. I just liked the basic Perl motto: TIMTOWDI. - Deepak Singh

worse is better - Bosco Ho

Uh Oh. This is going to get ugly - Deepak Singh

Once every ~6 months I convince myself that I should drop perl but every time there is something that needs to be done for yesterday and puff there goes the motivation. Its so hard to change when the transition is a period of veryyyyy annoying inefficiency - Pedro Beltrao

Pretty sure Ewan Birney popularized this; I remember it along the lines of "the first one to code it wins." The idea being that you can discuss object models and the right way to do things all day, but in the end working, easy to use code is what people are going to adopt. I don't remember if there is an official description of the thought process; probably too busy coding. - Brad Chapman

Unfortunately, this hides the idea that code is more than the an algorithm. If you had some clearly written, well documented, simple and well designed code which a new version of blast had broken, and a pile of spaghetti which happened to have been written more recently, perhaps it would make sense to dump the latter and fix the former. You can go too far with perfect design, of course, but you can go too far with macho programming as well. - Phil Lord

I think Ewan was the originator - I think I was just restating the motto. There was some design/developer philosophy in the bioperl.pod that also probably included this. The philosophy is that lots of mailing lists (where our dev is primarily coordinated) are filled with people spouting or complaining about things not being perfect, but in the end working code would win the argument. - Jason Stajich

I spent the day reading the source code to figure out the TRR format in GROMACS. It's an XDR format but you need to know the layout of the fields, which can be worked out by reading the write sequence in the C code. - Bosco Ho

They just recently released XDR format reader. Beats the good old gromacs template.c . PyMACS is nice. - Mitchell J Stanton-Cook

Can't seem to access the pymacs page. Is it a standalone library or a wrapper around GROMACS? - Bosco Ho

It's a wrapper. If I remember correctly it will handle .edr .xtc .trr and .tpr. It's incredibly easy to therefore build nice novel analysis programs. - Mitchell J Stanton-Cook

That last URL was a poor example; requires a whole bunch of Windows libraries on the machine. - Neil Saunders

Apache POI provides Java libs for Excel files. R is also an option - Rajarshi Guha

@shwu I've used xlrd in the past to read Excel data from collaborators, and xlwt to write it again - Michael Kuhn

@deepak http://www.resolversystems.com/product... - Piotr Byzia

Piotr, that's the one - Deepak Singh

Here's another one (never tried it myself tho): http://www.lexicon.net/sjmachi... - Bosco Ho

combinatorial chemistry? - Deepak Singh

evolutionary psychology? - Meg v. Meg v. 1.0.0.1

this? - Eric Jain from iPhone

hype cycles ... http://pbeltrao.blogspot.com/2007... - Pedro Beltrao

you need to update that. Personal genomics is way up on the hype cycle :) - Deepak Singh from IM

well it was in 2007 :) .. I should update it one of these days - Pedro Beltrao

systems biology. (That said, these things always work in cycles - the overhyping, the cries of false promises, the retreat from the limelight, and the slow integration into the mainstream) - Chris Miller

Chris, you have to look at Pedro's diagram then - Deepak Singh from IM

stem cells (oh snap!) - Wladimir Labeikovsky

Grand Unified Theory, String Theory - Sriks7

Super conductivity - Sriks7

was string theory ever a fad? High Temperature superconducting, absolutely - Deepak Singh from IM

Genome-wide assoziation studies, fMRI - Björn Brembs from iPhone

leeches? - Mike Chelen

I don't think HTS, GUTs or Strings can be called fads because they have been around too long for that. There are fads *within* those subjects for sure, but I think something that is popular for >20years is not a fad. - Matt Leifer

fractals - Jason Miller

chaos theory as a Rosetta stone for nature - Jason Miller

Nostratic. Though I guess it was never that popular. Generative grammar, then. - D0r0th34

Prions. - Neil Saunders

wearing sandals in lab - Mike Chelen

microarrays - Mr. Gunn

snap-bracelets - Mike Chelen from IM

+1 Jason. Chaos and fractals. Also cellular automata (though all 3 are very interesting fads) - Ian Holmes

Here's another good one that may be a bit more obscure: the catastrophe theory that was prostheletized by Zeeman. (Now, the mathematical theory of catastrophes is well established. The fad was to apply it to everything.) - Jason Miller

Wow. nice thread. Now..muhahaha... you have helped write half of a new blog entry. TYVM. Especially Pedro. - Iddo Friedberg

YOU_NAME_IT@home projects... or a bunch of crappy PCs contributing 24/7 to global warming. - Martin Jambon

Not sure that some of these things are fads. I guess that the ultimate test for phenomenon being a fad is that it ends up being discarded, or adopted at a a much lower key than it was initially played with. Maybe a good test would be to look for historic NSF/NIH/DOE roadmaps, and see whether the highly funded projects and topics of a decade or two ago actually "made it". - Iddo Friedberg

Open X Open Y Open Z - Abhishek Tiwari

yes i forgot about catastrophe theory! nice one. Can we add category theory too? - Ian Holmes

A bit outdated, but what about "solving the protein folding problem" or discovering a protein with a "novel 3D fold". - Mickey Kosloff

Mickey - Your 'protein folding' wasn't such a fad as a 'much harder problem' than people thought. Many people are still working on it, and it will have a huge impact when solved. - Jason Miller

I do maintain that "it's not rocket science" should be replaced with "it's not protein folding". Getting a ship to the moon's a lot easier problem to solve :) - Deepak Singh

@Jason - my personal view is that current state of the art structure prediction approaches (which is what most people are working on) are not what people *used* to refer to as "solving the protein folding problem". More specifically, before the CASP experiments started people used to frequently publish papers that stated that they solved this problem (again). My impression is that CASPs seem to have killed that particular fad. - Mickey Kosloff

Mickey, not the protein folding community (which is different from the structure prediction community). What happened was structure prediction got good enough and CASP (which I don't really like) gave people bragging rights, so that's where the effort went. Some day people will get back to real physics :) - Deepak Singh

@deepak. What happened with CASP was that a lot crystallographers and biochemists got sick of theorists claiming that they had solved the protein folding "in principle", which would happen like clockwork every few months in the 90s. CASP was put together as a challenge to put your money where your mouth is. Unsurprisingly the claims of having solved the protein problem dropped significantly after the first CASP. - Bosco Ho from iPhone

@ Deepak. I agree with Bosco, and perhaps I wasn't clear enough in my initial comment. The fad I referred to was not studying protein folding (either experimentally or computationally), but the (frequent) declarations of solving the protein folding problem. This was before my time, but people that have been involved in CASPs since the 90s still remember this "fad" well. I participated... more... - Mickey Kosloff

From my recollection most of the physical folding people never even took part because all you could even try and look at were small models (people like Peter Wolynes and Dev Thirumalai). It was the comparative modeling/threading folks who were participating and trying to out CASP each other with entire labs shutting down for half the year. That's just bad science. I think the first few CASPs had a purpose, and with all the structures being solved made a lot of sense, but CASP outlived its use in about 2004. - Deepak Singh

Now we're hijacking Iddo's thread for some CASP-bashing, which might require a separate thread (if not a separate room) :-) - Mickey Kosloff

... lol true. Sorry Iddo - Deepak Singh from IM

Work published in Cell, August 21, 2009 (after 2 years) - Ruchira S. Datta

Boltzmann visited Stanford in 1905. Gave first prescription about the protein structure, i.e., solving the folding problem: integrate the Boltzmann function. However, need to know entropy and enthalpy in excruciating detail. - Ruchira S. Datta

Also, proteins are not random polypeptides: they have been selected so they can fold quickly. Thus the energy landscape is a funneled landscape. - Ruchira S. Datta

So need to go back and forth between Boltzmann (physics) and Darwin (evolution). - Ruchira S. Datta

Function is an associated property of selection. Issue 1: correct level of selection--gene(protein)? cell? organism? group? species? Issue 2: history vs function, i.e., historical accidents vs functional necessities. - Ruchira S. Datta

E.g., serine protease S1 family has 1500 known examples. E.g., trypsin, chymotripsin. >1039 papers on this family in 2007. Binds serine of another protein in small pocket and cleaves it. - Ruchira S. Datta

Serine proteases function: catalysis and specificity. Hydrolysis of Suc-Ala-Ala-Pro-X-AMC. Chymotrypsin specific for X=Phe, Trypsin specific for X=Lys. But experimenters mutated trypsin to have specificity like chymotrypsin. - Ruchira S. Datta

Proteins subject to new Moore laws: number of available sequences increasing exponentially. - Ruchira S. Datta

History is reflected in sequence. If proteins are in same tree, will have historical and functional correlations reflected in the alignment. - Ruchira S. Datta

Can measure conservation. Conservation at each position can be measured e.g. by relative entropy. Conservation is heterogeneously distributed. The most conserved positions are in the core or at the functional interface. - Ruchira S. Datta

Could approximate conservation by checking whether amino acid is the most frequent or not. - Ruchira S. Datta

# Reminds me of our group's INTREPID: http://bioinformatics.oxfordjournals.org/cgi... - Ruchira S. Datta

Ranganathan: statistical coupling analysis (SCA). Look at conservation at *two* positions. Have covariance matrix and thus can check correlated conservation. - Ruchira S. Datta

Strong correlations are sparse and heterogeneously distributed. - Ruchira S. Datta

Diagonalize the matrix. Look at the eigenvalues i.e., perform spectral analysis of correlations. This is how correlations are usually studied. E.g., physicists studied correlations between stocks' time series in the S&P 500. Potters, Rouchaud, et. al. found that the eigenvalues corresponded to distinct sectors of the economy: transportation, paper, drug manufacturing, etc. - Ruchira S. Datta

Correspondence need not be one to one, e.g., a linear combination of eigenvalues could correspond to a financial sector. - Ruchira S. Datta

This analysis (PCA, not sure why he's not using this term) defines 2-3 sectors of the serine protease family. - Ruchira S. Datta

Are the sectors statistically independent? To test this, check correlation entropy: subtract the sum of the entropies for each sector, from the entropy of the two taken together. Found sectors are statistically quasi-independent. - Ruchira S. Datta

Project the sectors onto the 3D structure? Find that sectors are contiguous 3d substructures, which do not follow the secondary structure. - Ruchira S. Datta

Each of them occupies about 10% of the total sequence. - Ruchira S. Datta

found one of the sectors involved in catalytic site, another in the specificity swap, the third unknown from literature - Ruchira S. Datta

Halabi did alanine scanning mutagenesis of the rat trypsin. Plotted and found residues in one sector changed specificity without stability, whereas another changed stability without specificity. - Ruchira S. Datta

a very cool result for me is the 3d correspondence and some of the hints on their importance to function - Pedro Beltrao

Found strong cooperative epistasis using double mutants. - Ruchira S. Datta

Thus sectors have quasi-independent functions. - Ruchira S. Datta

Different sectors can separate: different subfamilies; or vertebrates and non-vertebrates; or enzymatic and non-enzymatic. Thus sectors are selected by seemingly independent evolutionary pressures giving quasi-independent functions. - Ruchira S. Datta

Thus besides the primary, secondary, and tertiary structure of a protein, can think of sectors as the functional structure. - Ruchira S. Datta

Functional sectors seem to be selected quasi-independently, i.e., as far as "levels of selection" we should go even below the level of protein. They were able to separate the historical vs functional correlations. - Ruchira S. Datta

Want to extend to other proteins, and protein interactions. Further work on mathematics behind correlations: best weights? how many examples needed? could have "pseudo-sectors" due to sequence weighting issues, coming from historical rather than functional correlations. What is the physics behind these 3d substructures? How many sectors per protein? Do they need to be independent? - Ruchira S. Datta

10th ICSB: Happy Birthday! - Ruchira S. Datta

one of the follow ups they are thinking about is similar analysis on protein interactions. that should be interesting - Pedro Beltrao

Have looked at some other protein families, e.g., SH2, SH3. Another group in Rockefeller has found sectors in another family. Interpreting the sectors must be done case-by-case. - Ruchira S. Datta

Would like to extend to pathways, but often don't have well-aligned sequences for all members of the pathway. - Ruchira S. Datta

Why is stability of serine proteases different in vertebrates vs non-vertebrates? A lot of work to be done. - Ruchira S. Datta

i wonder if they could use distance in protein 3d space to weight the correlations instead of conservation - Pedro Beltrao

I believe they should. It's insufficient to use conservation only. You need to use geometric constraints. http://www.ncbi.nlm.nih.gov/sites... - Deepak Singh

I spent all day trying to implement SCA in python. They have some explaining to do. Some formulas are rather mysterious, and files that they say are available on their website, are not. The results don't look too good. - Bosco Ho

what do you need Paulo? - Ian Holmes

Trying to do with DNA, what I did with proteins 3 years ago: http://www.biomedcentral.com/1471-21... - Paulo Nuin

cool -- our simulator is at http://biowiki.org/SimGenome -- see also http://www.biomedcentral.com/1471-21... - Ian Holmes

let me know if you're interested -- the tools should hopefully be self-documenting, so you may be able to use them independently, but i'd be happy to help - Ian Holmes

looking forward to it. Your protein paper was fantastic. Helped me settle on MAFTT. - Bosco Ho

Thanks! I saw Ranganathan give a talk on this a few months ago, have been waiting for the paper since. - Richard Klancer

No prob -- I think it's a pretty obvious idea, but it's good to see someone take the effort to do it and show that it's interesting. For me, actually providing the code is the most useful part. - Donnie Berkholz

Yeah, I've been waiting for this paper for a while. I also saw a talk a couple of months and it was absolutely fascinating. The SCA itself may be obvious, but the biological analysis and experiments were not. The code is indeed incredibly useful but I wish that it weren't in matlab and GUI orientated. - Bosco Ho

+1 Björn. @Pedro, you might want to have a look at the PLoS ONE blog. I think you may find this post of interest:- http://everyone.plos.org/2009... - Graham Steel

I think it's the best journal in the world today and likely to be recognized as such over the next decade or so. The combination of review for rigor not subjective qualities and willingness to explore post-publication metrics put P.ONE head and shoulders above any other journal, imo. I personally would be happy never to publish in any other journal ever again, because I could simply ignore IF etc and feel secure that my work would be available to whoever wanted it and able to be judged for its own merits. - Bill Hooker

+100 Bill - Graham Steel

^2 Graham - Peter Binfield

/ 0.0 Peter - Bosco Ho

Wow, what an incredible comment thread. - Kevin Gamble from iPhone

Well that's very beautiful and I agree, but I'm talking about real life now. I know the hate/love relation scientists have with impact factors and I don't think PloS One will excel on that matter, maybe also because of the quantity of articles published. Let's compare for instance with Journal of Biological Chemistry that also publishes a lot, but is an already very established journal in its area (traditionally). For a scientist/phd student at the beginning of its career is always a dilemma... - Pedro Matos

Pedro, perhaps you can take a look at http://www.int-res.com/abstrac... and start arguing with your supervisors from there. Or from http://www.slideshare.net/brembs... . Or from http://www.mathunion.org/fileadm... . I also dwelled a bit on this subject at http://ways.org/en... and http://ways.org/en... . - Daniel Mietchen

I also recommend this excellent post: http://shirleywho.wordpress.com/2009... - Peter Binfield

Thanks. I was also trying to play devil's advocate to bring on more discussion ;) - Pedro Matos

I agree with the nice things said about PLoS ONE in this thread, but the widespread community perception remains that this is a journal for all the stuff that didn't get accepted by "real" journals. I think there is a genuine dilemma for young scientists thinking of publishing there; they may very well be better off (in a ruthlessly pragmatic career-driven sense) finding another venue for publication, at least until PLoS ONE achieves wider recognition. - Daniel MacArthur

That makes three of us (another one is a commenter on Blind.Scientist) - Paulo Nuin

Well, I apologized to two of you already - Bosco, sorry if it didn't meet your expectations. Please keep in mind that we're still only in beta 0.6.3 - we're working on both speed and extraction quality. Beta 0.6.4 (next week) should almost double the extraction speed, beta 0.7.0 (in roughly six weeks) will use more external databases (e.g. PubMed) to improve extraction. - Victor / Mendeley Team

No apology needed. Just keep working on it. One disappointed "customer" is not the end of the world. - Paulo Nuin

I've been pretty harsh on some bookmarking services over the past years myself, so I understand how hard it is to please early adopters like us. - Mr. Gunn

I was testing two versions and both constantly crashed during indexing (in random places, so it's not the matter of a single pdf with errors). As Paulo says, maybe one day. - Pawel Szczesny

Well, I must have realy low expectations then, because I like Mendeley :-) - Björn Brembs

To be fair, mendeley is competing against endnote citeulike zotero and papers. The only way for mendeley to steal users is to provide the killer feature the others don't have and that is afaik pdf meta-data extraction. - Bosco Ho

I don't understand why they try to rely so much on the metadata extraction from the PDF itself (I might be wrong). It would be much easier, maybe slower but easier, to extract the DOI from the PDF and then get the metadata and PDF text online. The program already spends a good amount of time computing things, why not do it by accessing the data online? In my case I tested with 500+ files, I wouldn't mind if it took a couple of hours getting the data online, but eventually getting the information correctly. - Paulo Nuin

The Zotero v1.5 beta does have a PDF metadata detection feature, although I haven't used it myself yet so I don't know how well it works: http://www.zotero.org/blog... - John Dupuis

Paulo, I believe that's what Victor was talking about doing in his comment above. Bosco, I don't think it's a Zotero or citeulike vs. Mendeley situation. Mendeley tries to work more like an "itunes for PDFs". When you've already got an existing collection of PDFs on the desktop, it's not easy to convert that into a library in a online reference manager. It's also nice to be able to do a full-text search across your library and share files instead of just links. - Mr. Gunn

In that respect, Mendeley stands to get more converts from Endnote than from citeulike or Zotero, and that's something that I think we can all support, right? That's my motivation for working with these guys. I sought them out, they didn't approach me, and I did so because, although they didn't even have the bookmarklet out yet, I think they're closest to having the total package. - Mr. Gunn

I love Zotero and they've really done a great deal to force innovation on the rather stagnant field of academic reference management, but I just have a better feel for the direction in which Mendeley's heading with the recommendation and sharing features, and they're moving fast. Also, I guess I just prefer a desktop app for some things. I don't like a web service or firefox addin for... more... - Mr. Gunn

Apologies for re-hashing this thread, but FWIW, Zotero stores everything locally, too, and by default nothing remotely. I like it because I can use it offline (even though it's in a web browser). Also, since I search for papers using a browser, and Firefox, specifically, I've accepted that it's a browser plugin. - Chris Lasher

fundamental issue is recognition of interdisciplinary researchers, currently not good at all. - Abhishek Tiwari

I don't know they should study per se. Mastering your own discipline is hard enough, but they should be exposed to it, and IMO, a certain level of mathematics and statistics is a must - Deepak Singh

it's a complicated thing: whilst it's nice to make everybody learn a bit of everything, something has to give. In my experience, the vast majority of physics graduates never learn basic statistics. I've had to learn them on the job, even at various times, consulting psych statistics first year statistics. in physics, we used to do a service course to teach quantum mechanics to engineers... more... - Bosco Ho

Should biologists study computer science? No. Should they be aware that some of the principles of computer science - i.e. how to acquire, store, process and present information - are extremely useful for any scientist? Yes. Should they strive to improve their IT skills continually, over the course of their careers, not just in a "this is what I need to pass my course" fashion? Yes. I think we need a shift in how people learn, not what they learn. - Neil Saunders

Computational thinking perhaps? Maybe not at that scale, but the ability to think analytically as biology and other disciplines become more analytical? - Deepak Singh

here are a few links I used in a blog post http://boscoh.com/protein.... - Bosco Ho

Bosco, thanks a ton - Deepak Singh

svgfig / cairo - Michael Kuhn

reportlab - Paul J. Davis

http://people.csail.mit.edu/rasmus... - Michael Kuhn

http://nodebox.net/code... - Michael Kuhn

i use mostly matplotlib, but also the python imaging library to do some finishing touches like cutting and pasting, cropping, imagine resizing, text, and basic drawing elements. - Bosco Ho

Thanks everyone . I have started playing around with the matplotlib examples. What I want to generate is a 96 well plate with a color histogram or bars , to reflect concentration of two to three crystallization reagent concentrations . Overlayed on this graphical representation would be associated concentrations in numbers. - Hari

I guess you can get some idea with recent post of Rajarshi Guha where he implemented Plate Well Series Plots in R http://blog.rguha.net/?p=388 - Abhishek Tiwari

While we're on the subject can anyone point me in the direction of some matplotlib code or examples that output directly to a graphics file (i.e. without plotting to the screen). Thinking about making a simple web service for some data fitting. Actually if there is a nice example of using matplotlib to plot into a rendered screen in AppEngine that would be very cool as well. Or a list of curve fitting type services built on AppEngine. - Cameron Neylon

@Cameron, basically you need to use a backend that doesn't require X. Example http://helpful.knobs-dials.com/index... - Rajarshi Guha

Really hoping that someone has written a class that can print out reports for 96 well plates . Really hoping for something that gives fine grained access like grid(A,12).setbackgroundcolor(red) - currently struggling with layouts and font sizes and such in reportlab . - Hari

Why not display a heatmap type image of a matrix representing the plate (http://pyweaver.worldhoppers.org/api...) - Rajarshi Guha

yes indeed, it has a fun way of looking at the psychological theory and practice of deception. been a big fan of Tim Roth ever since Resevoir Dogs and Pulp Fiction - Mike Chelen

It's a great show. The Tim Roth character is allegedbly modelled after this guy: http://www.paulekman.com/about-e..., former professor at ucsf. He might even be my favourite fictional science nerd. - Bosco Ho

oh cool, I have to check that out. thanks Bosco - metalerik

I'm with the Zotero+Citeulike combo. - Paulo Nuin

Not sure what you're doing, but it's working fine for me. - Björn Brembs

It's true that the performance on a Mac isn't as good as on the other platforms, but at least it's trying to be cross-platform. Even on windows, I got a crash one when I first installed the new version, but nothing since. I just used it to format the bibliography for my dissertation, and there were a couple minor issues with that, too, but nothing show-stopper. For the previous paper, I used Zotero, and they're pretty much equal in their Word integration, IMO. - Mr. Gunn

It was even worse on a quad-core Vista machine. Not that Vista is that good too, but ... - Paulo Nuin

In my work on Avogadro Vista causes the most cross platform problems. It will work on XP and even Windows 7 RCs... The Mac Qt builds have certainly had some issues too, but things are improving. - Marcus D. Hanwell

It works fine for me on a win vista but I have it with only a hand full of papers. no crashes and decent speed. There were some weird bugs in the online version but that could be Chrome that does not display pages well on my current resolution. - Pedro Beltrao

For comparison, I just downloaded the latest version of Papers. It doesn't get past rebuilding the database before crashing. - Bosco Ho

I tried it recently but it couldn't seem to sync with CiteUlike. I tried clicking the sync button a couple of time over a few days but no luck. - Michael Barton

I'm "retesting" Papers at the moment with the same input files I used with Mendeley. The import is taking about the same amount of time because I selected the option to copy the files to its own DB. But the amount of used CPU is far less than Mendeley. - Paulo Nuin

It constantly crashes on my Mac,about the only thing that does apart from MS Office, boo! Sticking with Zotero, Connotea and Delicious for different types of bookmarks. - Sally Church

Metadata extraction isn't that efficient, pdf rendering is still slow, and it fails to recognize many (almost-)duplicate entries, but I liked the new improvements. It is not ready for productive use, though. This version did not crash on linux, but sync stopped working a few days ago. I was told they were changing servers and that syncing won't work for a while. - Bruno C. Vellutini

Paulo, I agree with most of what you say in your review (e.g. the not-so nice Mac interface). It's still beta software. But I see the improvements over time, and I'm optimistic that this will turn into a fine program. - Martin Fenner

I agree with what Martin just wrote, plus it's hard to knock free-to-use software that has obviously had a lot of effort put in. - Michael Barton

Let's not forget that Mendeley is not a charity. It will be a paid application soon. - Paulo Nuin

Victor says the application will remain free. If you look at the business model of last.fm (with which Mendeley shares board members) you can get an idea of what they're thinking about re: income. - Mr. Gunn

Yes, right. - Paulo Nuin

Mendeley desktop was constantly crashed on my PC, after update for last version seem like fixed but I barely used after that so can't conclude. Connotea is always very slow, frequently down and doesn't work with ScienceDirect -it's getting annoying. Happy with Delicious. Can't afford Papers. Thinking to try Zotero. - Alexey from iPhone

IMHO, Zotero is the way to go here. - Paulo Nuin

Zotero is a fine application with some great development talent behind it. Can't say enough good things about those guys. - Mr. Gunn