Sign in or Create an account
Matt Wood's Comments - View full feed
FriendFeed
Matt Wood posted a link
4 minutes ago - via Bookmarklet - Link
Embrace the fail. - Matt Wood via Bookmarklet
Twitter
Pierre posted a message on Twitter
Twitter
Deepak posted a message on Twitter
Google Reader
PauloNuin shared an item on Google Reader
7 hours ago - Link
And for the Rubyists, take a look at Skynet: http://skynet.rubyforge.org/ - Matt Wood
Twitter
Roger Pettett posted a message on Twitter
Twitter
Neil Saunders posted a message on Twitter
FriendFeed
Thursday at 5:57 am - Link
Enabling! I'm so done with doing research right now. Too slow, too frustrating. It's fun now that I'm writing up my thesis, but every chapter was two years of work and I was unhappy and stressed and suppressed 99% of my time at the bench. - Eva
doing research. But mother Nature said "NOOOOO!!!!, the bench is not a safe place for Science when you're using your pipetman" :-) - Pierre
enabling. I like the variety of challenges that brings as opposed to the microfocus of research - Daniel Swan via twhirl
Doing it, definitely. I've been a patch clamper a long time now, but I still get a thrill when I see currents up on the screen. There's nothing like it. - NatBlair
I loved doing research myself, but as a career, definitely the latter. - Deepak
Enabling: adding value to anything which helps the scientific method. Always different, emerging and interesting. - Matt Wood
Enabling it. I have the same feeling as Eva and Matt. - PauloNuin
both, but like to walk on the computational side - Attila Csordas
occasional bench research + running lab as a PI + teaching students - 2/3 of the current hellish grant burden = perfect - Maureen
enabling. greater variety, plus I still get to do all the original thinking I want to. Also I was a danger with P32. - Simon Cockell
I love doing research, but I hate writing it up. I enjoy writing well-crafted Ruby code to answer scientific questions, but I'm utterly useless at communicating the results. - Michael Barton
Haven't had much experience enabling yet, but I have the feeling I'd enjoy enabling research more consistently than doing research. But doing research does give a thrill if it goes well and you can keep the interest alive. - Shirley Wu
I do lots of "research" -- on the internets... - Eric Jain
Enabling & Connecting Minds. In doing my focus seems to move towards empirical work, statistics, anything in highly complex macro-systems. Ah, science in industry is cool. - dekay
It used to be research but then it got to a point where there were no tools available that could meet my needs, so now it has turned to enabling and trying to develop those tools. - Paul Bacchus
maybe I'm on the enabling side? Well fact is I've developed my share of computational methodologies and software so far, and it was fun!, but when it comes to thinking about the biological theories/hypotheses/paper writing from the data the software spit out, I kind of hate it. So I guess I'm no into "doing research", but more for hacking the heck out of it... - Ntino
I just don't know. I'm having a really hard time with that exact question right now. I'm nearing my use-by date as a postdoc and not sure I really want to try to scramble up the faculty food chain. - Bill Hooker
I'd say 60:40 - Ricardo Vidal
@Bill - hard time with that exact question is precisely my feeling and reason for the post. - Neil Saunders
the answer is "translate the the clinic!!!!!" not from bench to the bench - Alexey
Google Reader
Deepak shared an item on Google Reader
Wednesday at 4:00 pm - Link
err ... folks .. better start re-writing your code now - Deepak
parallelizing legacy code is a tremendous challenge; the tools suck and we aren't really training future programmers to do it; what will happen? - Adam Kraut
I thought we were hitting the limits of silicon semiconductors. What happened to "optical/quantum computing is the future"? - Neil Saunders
Having worked in the molecular electronics field in the past, I've heard the "this is the future of computing" all too often. Definitely fall in the "silicon has more legs than you think" camp - Deepak
Good overview of some of these threads from Pat Helland: http://is.gd/L5q - Matt Wood
One core per Twitter clone.. - Nick Lothian
Blog
Wednesday at 10:04 pm - Link
Couldn't agree more. Love it. Should we wait for someone, or just start pushing the boundaries? - Matt Wood
I was just about to get around to that blog post about SNP profiling being a disruptive technology. Was even going to use that phrase, too. We must be reading the same people, Deepak! - Mr. Gunn
Neil and I have made a habit of telepathic blog posts. I should add that for a moment I was tempted to put a picture of a Klingon disruptor there :) - Deepak
I'll add that the technological disruptions are there (I was pretty much focused on SNP profiling in the clinic for the past six months), but how people use those technologies. Also, even with SNPs, the actual clinical utility is still limited (see Eric Schadt's work on needing network analysis etc to supplement GWAS) - Deepak
I've got into a habit of no blog posts; hopefully that will change on return from ISMB. So other than "just go for it", how would someone (practically) begin to drive a disruptive process? Let's say my goal is "revolutionise drug development". Academia is out - such a grant will never fly. So I need to think startup. Write a business plan, win over a wealthy philanthropist? - Neil Saunders
If I knew, I'd be on a beach somewhere :) - Deepak
FriendFeed
Matt Wood posted a link
July 1 at 10:35 am - via Bookmarklet - Link
Making data portability suck less, apparently. - Matt Wood via Bookmarklet
Tumblr
Matt Wood posted an item on Tumblr
July 1 at 5:54 am - Link
Hey, Matt, if you are talking on FF and Twitter on the conference you should set up a FF room for the conference to prove it! :))) - Attila Csordas
I can only prove my presence on the conference if I call it the Science (micro)Blogging 2008 conference. :) - Attila Csordas
Attila - excellent idea! http://friendfeed.com/rooms/sc... - Matt Wood
SlideShare
Matt Wood posted a slide show on SlideShare
Genomes On Rails
June 25 at 11:44 pm - Link
A repost due to the World's Best Presentation competition, now on over at Slideshare. - Matt Wood
Twitter
Matt Wood posted a message on Twitter
Twitter
Neil Saunders posted a message on Twitter
Twitter
Neil Saunders posted a message on Twitter
FriendFeed
June 20 at 3:17 am - via Reshare - Link
Is anybody planning to attend this conference? - Martin Fenner
I'll be there: was also thinking of arranging some kind of scientific software summit around there, too. - Matt Wood
I went last year, but haven't decided on going again. But I would be around to do something. - Martin Fenner
del.icio.us
Jeremy Zawodny bookmarked a page on del.icio.us
June 19 at 7:20 pm - Link
I saw this coming a mile away... - Jeremy Zawodny
This is what they call an exodus - Deepak
Joshua always copies Flickr! :^) - Christian Crumlish via Alert Thingy
For me this one is particularly sad. del.icio.us remains my most used web service, and despite Yahoo neglecting it for so long, I've never found anything to replace it. - Deepak
My question is: what's going to happen to my Flickr, del.icio.us data? - l0ckergn0me
It is all looking a bit ominous. I like del.icio.us, but there are very similar alternatives to turn to. Flickr is a different story; IMO it's by far the best photo manager out there. - Neil Saunders
@l0ckergn0me : read yahoo's TOS :) - directeur
... delicious isn't being neglected by Yahoo. - Dave Dash
Excuse me while I go and back up my bookmarks - Adewale Oshineye
@Dave - as a user of delicious all the publically available evidence points to it being completely ignored... - Andy Davies
Del.icio.us could have become anything if Yahoo had put resources behind it. You can see the genes of delicious in things like Digg & FriendFeed now - but delicious itself has gone nowhere. - Nick Lothian
Delicious has in a way fulfilled its historical mission: spreading the tagging meme. - benedikt
Good call benedikt!! Tags as touching points born on del.icio.us, next stop mento.info ;-) - cbgreenwood
Nick Lothian is right. del.icio.us could have become something huge. but they went nowhere with it. it's pretty unbelievable actually - marcel weiss
another sad day on the internet. - andre
Blog
June 18 at 2:49 pm - Link
If it is possible to have DOIs for objects (or, so they say, enough IPv6 addresses for every molecule on Earth), why is it so difficult to implement DAIs for authors? - Duncan Hull
Because everyone wants their own ;-) - Cameron Neylon
ask NCBI... - Pierre
I'm not sure NCBI see it as their problem: although they are clearly best suited to fix it. There is Spock, but that doesn't seem to have gathered much traction: http://www.spock.com/ - Matt Wood
I think biomednetwork has probably implemented this the best - at least in terms of figuring out who is who - but all of these things seem to be limited to abstracts in PubMed which is a bit limiting really. Need something that publishers actually want to implement and which is an 'opt-in' system. And who gets to assert who the authors on a paper are - the corresponding author - the poor student actually doing the submission? - Cameron Neylon
This is one of those issues that everyone agrees on and that comes up once in a while but for some reason it still is not implemented. NCBI would be the best place to start. Is there any argument against an universal author ID number for science ? The only questions would be when does someone get recognized as an author ? Who certifies the authorship? - Pedro Beltrao
I think all the efforts will be vain as long as pubmed/NCBI , the leading player, will not add an author-id in the pubmed records. It needs a database of author with their name, their location, their... hey... we already suggested this yesterday, didn't we ? :-) http://friendfeed.com/e/5a1a4d... - Pierre
The technology is there to do it http://en.wikipedia.org/wiki/O..., I think the main barriers are sociological and political. The NCBI is not necessarily the best place to start, like DOI's, it needs to work outside (as well as within) biomedical and life sciences. If you like commercial companies being in charge of your identity then there is http://www.researcherid.com, http://www.scilink.com and hundreds of other people eager to take on your digital identity for you... - Duncan Hull
Online identity is probably the next big thing on the web. Whether it is you as an individual, or you as a scientist, or you as a furry, there needs to be some control of who you are and what you can claim. Maybe the future is some kind of marriage between openid + researcherid type system, i.e. one where you can claim a domain specific id, via some global ID system. - Deepak
This is actually kinda a gnarly thing, when you think about it. Theoretically, ICANN or DOI could have a little side-business assigning identifiers, but do we really all want permanent unchangeable serial numbers? - Mr. Gunn
Twitter
SciPhu posted a message on Twitter
Twitter
Pawel Szczesny posted a message on Twitter
Twitter
Neil Saunders posted a message on Twitter
Twitter
Matt Wood posted a message on Twitter
Twitter
Jan Aerts posted a message on Twitter
FriendFeed
June 17 at 4:48 pm - Link
clustering of what? - PauloNuin
Yeah, that was my question :) - Neil Saunders
Depends on what type of data you are looking at. Average and complete linkage are probably common, but I've noticed a lot of people move to other forms of clustering lately, especially SOMs (for gene expression data) - Deepak
I'd like to see a method that clusters all the different clustering methods out there :-) - Eric Jain
Meta-clustering. Well if we can have meta-programming, why not - Deepak
Well, nominally I'm talking about binary vectors, but it's possible I will work with non-binary vectors too. The vectors are features derived from sites in protein structures. Perhaps the answer is totally dependent on the nature of the data? - Shirley Wu
I just realized you work for Russ Altman. Didn't they publish a k-means clustering paper on protein microenvironments last year? Is there a reason you wouldn't use that? Going back to your original question, I suppose Pearson Correlation might be a good place to start and apply on a test set to see if you get some decent clusters - Deepak
K-means seems to be pretty popular, and is relatively easy to implement. From that, you should get some idea as to whether additional clustering approaches would be useful. - Matt Wood
Pearson is a good distance measure for variables in a dataset. An example would be clustering stats about a group of people such as height and weight, you cannot compare these directly, but you might be interested in how one variable changes with the other. Euclidean distance on the other hand would be useful for clustering the records (each person) based on their height and weight i.e. sum of the difference between height/weight of person A and person B, where you can compare variables on the same scale. - Michael Barton
You could also think about row/column normalising a data set to prevent over-variation biasing the clustering method. This wouldn't be applicable to a correlation based distance measure, but would be for a euclidean based distance measure. One possible way to normalise is to scale each row/column to have the same mean and variance. - Michael Barton
I realise I didn't answer your original question. I like centroid based linkage because it's median-based and is a little more robust to outliers. However this might not be as useful if the data is non-globular. As Deepak mentioned SOMs are good, as well as Neural Gas, and Growing Neural Gas. If you're using Matlab I think there's a good package with all these things in. - Michael Barton
If you're using R, Mclust is pretty good. You don't have to decide the number of clusters before hand, the model calculates the number of clusters in the data based on maximising a byesian information criterion. - Michael Barton
@Deepak, yes, we published a paper with k-means, and now we're going back to those clusters and would like to do a "better" clustering within them - idea being that k-means got us into a reasonable ballpark, now we want to have more refined clusters. - Shirley Wu
The last two days I've been using Cluster 3.0 (with TreeView to visualize) which offers probably half a dozen distance metrics and 4 or 5 linkage methods. I'll look into Mclust for if we decide to recluster everything, but we're talking 2 million vectors here - can SOMs or Mclust handle data on that scale? - Shirley Wu
Use Dendroscope for visualization. Better, more powerful. - PauloNuin
I've never used Cluster unfortunately. Two million vectors is large data set, I'm not sure how R would cope with that, I guess the only way to find would would be to give it a go if you have the time. As for SOMs I guess it would depend on which platform, I think Matlab should be able to handle such a sized dataset. - Michael Barton
Two million vectors in R would take quite a while and good amount of memory. - PauloNuin
FriendFeed
Deepak posted a link
June 16 at 8:55 pm - via Bookmarklet - Link
Coolness - Deepak
Funny ! here is mine ( phew ! Bioinformatics is the biggest word !) http://wordle.net/gallery/pier... - Pierre
I get "applet not started". Whatever. - Neil Saunders
Posts from Green is Good: http://wordle.net/gallery/Gree... - Matt Wood
pretty cool - Mr. Gunn
Can you say I am a biologist? http://wordle.net/gallery/nuin... - PauloNuin
Twitter
Matt Wood posted a message on Twitter
Twitter
Pierre posted a message on Twitter
FriendFeed
June 12 at 5:29 pm - via Bookmarklet - Link
hmm, is it all losing steam - Jason Stajich
Lincoln Stein's prediction is coming true :) http://www.oreillynet.com/pub/... - Neil Saunders
check the top country, India, and the longest bar is bioinformatics -- haven't searched the matter, but is there outsourcing of bioinformatics application coding there like with the other types of software ? - Ntino
They changed the scaling algorithms in the new version right ? I cant seam to make a query that results in an upward trend. Maybe buggy ? Alternative .. a lot more people getting online and therefore reducing the relative number of science related queries (its a relative measure). Or just even pople doing a lot more queries in general, most of them just to find a website, etc - Pedro Beltrao
Agree with Stein's writing... it's a tool, even though to do, but will become a field generated by the next stage in the evolution of the biology field... cause biology is becoming more and more data intensive. Do you think in 2012, anybody will be going to the lab to do cultures on Petri dishes and observe colonies changing colors (sorry wet lab people in the room). Sure there'll be labs, but Core labs with the data spitting instruments, like array scanners, sequencers, auto-protein gel scanners etc. - Ntino
... to continue the rambling: all data out of those machines sucked into computers (well we hope for e-lab notebooks, or better for the data in the open) and then that 's where the discovery will be... that's how I see the future of biology, which is currently happening, turning into a data-intensive discipline. - Ntino
Note that the news volume (bottom chart) has actually increased. The search stats tend to be a bit weird (except for very popular search terms, which bioinformatics is not), so I wouldn't use that information for investment advice :-) - Eric Jain
I sense debate brewing :) I think there will always be a place for people looking at things in labs. I like bioinformatics but I don't think it ever has been, is, or will be a field. The sole reason for bioinformatics journals is to give academic bioinformaticians a place to publish: because that's the sole output recognised by academia. Academia doesn't know how to employ tool-builders, or recognise their skills. - Neil Saunders
One reason "bioinformatics" collapsed as an industry was for some of the same reasons listed above. It was never an industry by itself, just got subsumed by the broader scientific process (by and large) - Deepak
To blame: Leona Lewis? http://is.gd/wAp ; ) - Matt Wood
To be honest, that was a joke: but what's going on towards the end of 2006? A big spike for Leona and a big corresponding dip in science searches! She _is_ to blame! - Matt Wood
:) I blame twitter, same observation. - Neil Saunders
Matt ... lol :) - Deepak
FriendFeed
Attila Csordas posted a message
June 9 at 11:47 am - Link
It's going to be a maximum of $199 USD in any country, according to Leo Laporte. With 3G and GPS builtin, impressive. - Chris Lasher
Remember me to tell you the price in Brazil after the release date. Everything there has at least 100% premium. - PauloNuin
I find it difficult to believe that there won't be a 100% markup for the UK market - Cameron Neylon
Chris, it's only impressive compared to itself. Newer model Nokias have had 3G and GPS and Wifi for years, and they're not any more expensive. You also don't have to pay $10-25 extra for unlimited data. Oh, and the screens are higher res. - Mr. Gunn
yeah, but that UI and the web browsing. I don't have one, but it's been cause of the price - Deepak
My Nokia E51 feels dated now. At least it's a single-handed phone, one of my decisions not to get an iPhone (or Blackberry). - Roland Krause
Deepak - Did you know that the N and E series Nokias (Symbian smartphones) use the same browser as the iphone? iphone sites/apps generally work on the Nokias, but you can also install local apps when you need speed, like for maps and gmail, etc. - Mr. Gunn
I've heard already approximate price in Poland for 8GB model - ca. $270 when signing two year contract. Still it's pretty OK, I would expect much more. - Pawel Szczesny
Mr Gunn - it's the same rendering engine, WebKit, rather than the same Browser, Safari. It's _all about_ the pinch zoom. ; ) - Matt Wood
I'll be surprised if they don't have a huge mark up in Australia. My prediction: 8GB=US$300, 16GB=US$400, only available on 24 month $50+ contracts (at least for the first few months while the hype subsides). Lets see how close my prediction is. I'll be thrilled if I'm wrong, since that means I may get one (after closely comparing the Nokias ... I agree with Mr Gunn [ http://www.thebestpageintheuni... ]) - Andrew Perry
LOL, Andrew, I thought about posting that, and I do have PuTTY on my phone. - Mr. Gunn
Blog
June 12 at 6:42 am - Link
Good luck at your new gig! - J. Phil
From somebody who's Internet experience has been heavily influenced by you (you were one of the first blogs I ever followed), best of luck and I look forward to seeing the results of your new opportunity! - Damon via twhirl
yahoo won't be the same without you.. :( - Augusto Becciu via twhirl
Just to hope you all good to your new work, whatever it is. - Daniel Schildt
Where are you going? Anywhere we've heard of? - Tony Ruscoe
Jeremy. Good luck. I can only echo what Damon said above - Deepak
good luck - Andreas Gohr
shocked to know that u'r leaving. good luck. - Jansen Lu via twhirl
shocked. And happy for you :) - John Mueller
woo. crazy. all the best to you. - Josh via twhirl
Congrats on a huge move, and all the best, Jeremy. You'll be deeply missed as a Yahoo, even by those of us that don't work there. - DeWitt Clinton
Welcome to the ex-Yahoo club, Jeremy, and congratulations on the move. Looking forward to seeing what comes next for you. - Ryan Kuder
Congrats! - Ginger Makela
Best of luck! - Mike Reynolds
Who's the next lucky one to hire you!? :) - directeur
Good luck! - Matt Wood
Reading this in Friendfeed. Doesn't seem appropriate to "like" it. More "ackknowledge that it is news". You seemed like a force for good in a good company Jeremy. - Scott Koon via twhirl
Congratulations ;) What's next ? Google or Own project ;) - David Berrebi via twhirl
@Jeremy startup please!! ;-) - Erhan Erdogan
He said a 'smaller' company. So that rules out Google. - Adewale Oshineye
Best of luck Jeremy ! - Charlie Anzman
Good luck Jeremy, can't wait to see what your up to next. - Barry Schwartz
Wow! All the best man! - Robert Cooper via twhirl
Wow! Good luck Jeremy! - Niranjan Tulpule
The problem with having an ! in your company name is people always sound so excited when they write about leaving - Christian Anderson