I guess that as a CAFA organizer, I should really contribute to the second page. And I will. But I really don’t mind if someone else jump-starts it.

http://en.wikipedia.org/wiki/Protein_function_prediction

http://en.wikipedia.org/wiki/Critical_Assessment_of_Function_Annotation

It is with no inconsiderable degree of reluctance that I decline the offer of any Paper from you. I think, however, you will upon reconsideration of the subject be of opinion that I have no other alternative. The subjects you propose for a series of Mathematical and Metaphysical Essays are so very profound, that there is perhaps not a single subscriber to our Journal who could follow them.

David Brewster, physicist and mathematician and inventor acting as editor of The Edinburgh Journal of Science to Charles Babbage, mathematician, philosopher, inventor and mechanical engineer; father of the computer circa 1821.

Hats off to Jonathan Eisen for hosting this activity on his blog. (I’ll keep mine on, thank you. It’s raining cats and dogs here right now).

A couple of weeks ago I posted a discussion about two papers that challenged the ortholog conjecture. Briefly, both papers stated that orthologs may not be such great predictors for molecular function. One study from  Indiana University by  has shown that paralogs may be better predictors than orthologs for molecular function. Or, at the very least, paralogs should not be excluded as predictors. This paper has generated quite a bit of interest and controversy. Consequently, Eisen has invited Matthew Hahn, the lead author to write about “the story behind the story” in Eisen’s well-read blog. The post is a great read, and has generated an animated discussion in the comments area. You do need to clear quite a bit of time to go through both Hahn’s guest post and the comment thread: the topic is a rather complex one, and as explained in the comments thread, one problem is that the ‘ortholog conjecture’ itself seems to be not well-defined.

I kept checking in to Eisen’s blog to read the elongating comment thread. It seems that now a special session on the topic may be in the works for the 2012 annual meeting of the Society for Molecular Biology and Evolution following this discussion. So great to see such an involved community getting together.

1. This is crap. (Rare).

2. This is terrific. (Even rarer).

3. This can be OK, but they really need to fix A, B & C. (fairly common).

4. If I only knew what they meant in point A, I could say whether they need to fix A, B & C or just B & C, or maybe explaining A will clarify to me that B need not be fixed at all.(Really common!)

The things is, sometimes draft manuscript are good, but unclear on a very specific and crucial point. Like point A.  The lack of clarity is not necessarily due only to bad writing. Sometimes it is very hard to describe a new idea or method. After all, it is new, right? So no-one has described it before and finding the correct way to describe you new idea or method can be very tricky. Some people use analogies (like me, I love analogies). But an analogy can be over interpreted, and then mis-interpreted, and your intricately concocted explanation is all shot. Then again, some authors do not use analogies, but explain their methods very formally. A reader who is very visual (like myself) would like to see a some graphic depiction of the method, but that may not be possible, or the author’s graphic rendering skills are not good enough.  In any case, I seem to get a lot into the position where my opinion of the paper hinges around one or two crucial details. If this were a seminar, and the author was talking about her work, I would just ask her to clarify the point. But I cannot do that when reviewing an article. Which leaves me with asking to perform major revisions, which sets the paper into a serious delay mode or sometimes reject because other reviewers have been less amenable.   (As an aside, I have noticed some reviewers  just let  the ambiguity of point A slide. This usually happens if the last author is Dr. Bigschotte, holder of the Endowed Golden Chair professorship of Biowizardry).

But maybe I can ask that clarification question of the author, or even have a brief dialogue? Anonymous email/chat technology can do wonders for shortening the turnover time of papers and clarifying issues. I am not saying that writers are now given a free license to write badly. But if needed, a chat session or email exchange monitored by the editor could really help push a paper through (or away). The exchange should be very brief, topical, logged (with the referee anonymized). The session should be requested by the referee, with very specific questions. The number of back-and-forth exchanges should be limited.

As a referee, I see myself more as a midwife (or whatever is the male counterpart) than a gatekeeper. I am not interested so much in keeping bad papers out (that is actually fairly easy), but letting good science in, even when it presents itself feet first and covered in gunk (OK, that was a rotten analogy, but you know what I mean). Anything that can ease this process is more than welcome.

So… any takers? Or is this a terrible idea? Better yet, has something like this been done?

PS: actually, as an author, I sometimes receive reviewer comments which I do not quite understand. This is even worse, since if I don’t address the issue in the way the reviewer asked for, it can spell the death of my paper. So this mechanism of quick dialogue between reviewer and author can work well both ways.

I am learning a lot about scavenging tweets.  Apparently, I cannot go back beyond a few days using the api.search() function. Hence, if I try to search for all the #AFPCAFA11 hashtags I will get nothing from the meeting’s dates. But if I look for a user’s tweets using api.user_timeline() I can go back for months on the users timeline, and then filter out the tweets with the relevant hashtags.  Since it seems I was the principal twiterrer in that meeting, I’m putting up my tweets here. Apologies to the others who recorded the meeting using Twitter: if you want your tweets included, drop me a line with your Twitter user name.

Thu Jul 14 16:04:22 2011 At Vienna. #ISMB #AFPCAFA11 #ISMB11
Thu Jul 14 16:07:34 2011 At Vienna #ISMB #AFPCAFA11 Curious who is the best protein function predictor? Join us. http://bit.ly/htv3J7
Fri Jul 15 09:10:24 2011 Jesse Gillis from UBC on a function prediciton post-mortem #AFPCAFA11 #ISMB
Fri Jul 15 09:12:11 2011 This is going to be fun. Jesse Gillis UBC. Postmortem on MouseFunc #ISMB #AFPCAFA11 Precision / Recall of 0.06… argh.
Fri Jul 15 09:14:14 2011 http://bit.ly/obSISi MouseFunc experiment #AFPCAFA11 #ISMB
Fri Jul 15 09:28:58 2011 Multifunctionality affect prediction profoundly. Take-home message from Jesse Gillis’ talk #AFPCAFA11 #ISMB
Fri Jul 15 09:31:22 2011 Next up: Meghana Chitale from @kiharalab #ISMB #AFPCAFA11
Fri Jul 15 09:36:44 2011 Co-occurrence association scores. CAS Lookiong for associations across GOs: between a BPO term and a CCO term, for example. #AFPCAFA11
Fri Jul 15 09:41:17 2011 Missing enzyme predictions. My fav. Chtiale at #ISMB #AFPCAFA11
Fri Jul 15 09:56:08 2011 Yanay Ofran from Bar ilan U about multifunctionality. How to assess the number of false positives? #ISMB #AFPCAFA11
Fri Jul 15 09:56:44 2011 Prediciton of photosynthesis in an elephant genome is a good sign of false positives. Yanay #ISMB #AFPCAFA11
Fri Jul 15 10:02:56 2011 Precision of short motifs is surprisingly high. Yanay, #ISMB #AFPCAFA11
Fri Jul 15 10:05:42 2011 short motifs identify functional motifs. Whereas homology identifies evolutionary relatedness. #AFPCAFA11 #ISMB
Fri Jul 15 10:09:14 2011 Next up: me. #AFPCAFA11 #ISMB
Fri Jul 15 11:44:45 2011 David Jones from UCL is talking about his #AFPCAFA11 predictions. Many different features. #ISMB
Fri Jul 15 11:47:18 2011 profile-profile fold recognition works well in Function prediction as well. #AFPCAFA11 #ISMB
Fri Jul 15 11:53:02 2011 Jones talking about why it is unhealthy to exercise in the morning. #AFPCAFA11 #ISMB generation of free radicals. #excusesaregreat
Fri Jul 15 11:56:43 2011 49,000 features in an SVM. #ISMB #AFPCAFA11
Fri Jul 15 11:57:47 2011 Hard to believe no redundancy in 49K features…. #ISMB #AFPCAFA11
Fri Jul 15 11:59:17 2011 “I like this plot and I would make a t-shirt out of it, but in terms of scientific value its worth is zero”. Jones, #AFPCAFA11 #ISMB
Fri Jul 15 12:01:09 2011 New term heard for a 2nd time at #ISMB #AFPCAFA11 “postdiction” as opposed to “prediction”. #notsurewhatitmeans
Fri Jul 15 12:16:49 2011 Lightning talks at #AFPCAFA11 #ISMB starting now
Fri Jul 15 12:35:51 2011 Mary Jo Ondrechen on SALSA at #AFPCAFA11 #ISMB structure –> function.
Fri Jul 15 12:36:45 2011 CDEHKRY make up 75% of all catalytic sites. Mary Jo #ISMB #AFPCAFA11
Fri Jul 15 12:50:50 2011 Jeffrey Yunes from UC Berkeley on SIFTER from Steven Brenner’s lab. #AFPCAFA11 #ISMB
Fri Jul 15 14:28:20 2011 Patrik Koskinen on a function prediction method called PANNZER. This will roll over well. #ISMB #AFPCAFA11
Fri Jul 15 14:30:30 2011 More information on PANNZER and the other methods at #AFPCAFA11 here: http://bit.ly/l9ayW9 #ISMB11
Fri Jul 15 14:33:16 2011 Koskinen mentioning Biothesaurus http://bit.ly/rnHdzp which removes errors due to synonyms #AFPCAFA11 #ISMB
Fri Jul 15 14:43:09 2011 Question: “How do you pronounce the name of that volcano that erupted in Iceland?” Answer: “I don’t”. Koskinen #AFPCAFA11 #ISMB
Fri Jul 15 14:50:27 2011 Olivier Lichtarge on using Evolutionary Trace Annotation (ETA) for function prediction. #AFPCAFA11 #ISMB
Fri Jul 15 14:51:12 2011 A network of protein structure networks. Memories of fragnostic. #AFPCAFA11 #ISMB
Fri Jul 15 14:52:49 2011 Using network diffusion to annotate protein structures. Lichtargee. #AFPCAFA11 #ISMB
Fri Jul 15 14:57:19 2011 compressing a clique to a star graph by adding a pseudo-node. Reduces problem from O(n^2) to O(n). Lichtarge #ISMB #AFPCAFA11
Fri Jul 15 15:09:17 2011 Amos Bairoch: of prosite, swissprot and expasy fame #AFPCAFA11 #ISMB
Fri Jul 15 15:12:51 2011 Due to alt-splicing and PTM 20,000 human genes –> 5M different molecules! Bairoch #AFPCAFA11 #ISMB
Fri Jul 15 15:15:50 2011 Status codes of human protein function annotations: Maybe, potentially, putative, expected and hopefullly. Bairoch #ISMB #AFPCAFA11
Fri Jul 15 15:17:10 2011 >100 GPCRs for which we do not know the ligand. Bairoch #ISMB #AFPCAFA11
Fri Jul 15 15:19:13 2011 Bairoch now talking about CALIPHO. 1)experimental verification of human protein function; 2)enable bioinformatics for same. #ISMB #AFPCAFA11
Fri Jul 15 15:20:44 2011 “How many ppl in this room have never used swissprot”. 0. #AFPCAFA11 #ISMB
Fri Jul 15 15:23:24 2011 Bairoch looks at small <100aa intracellular protiens in experimental assays. #AFPCAFA11 #ISMB
Fri Jul 15 15:26:26 2011 Interesting proteins are expressed in olfactory pits of zebrafish. I didn’t know fish smell. #AFPCAFA11 #ISMB
Fri Jul 15 15:33:09 2011 If you get a new function, you cannot predict it, because of no ontology (yet). #AFPCAFA11 #ISMB
Sat Jul 16 06:58:43 2011 Predrag Radivojac explaining the vagaries of GO annotated databases #AFPCAFA11 #ISMB
Sat Jul 16 06:59:48 2011 Assessment of protein function prediction methods going on now in Hall L #ISMB #AFPCAFA11
Sat Jul 16 07:04:08 2011 Radivojac: “It’s possible to achieve a precision of 1, it just won’t happen”. #AFPCAFA11 #ISMB
Sat Jul 16 07:29:50 2011 Assessment of protein function prediction methods going on now in Hall L #ISMB #AFPCAFA11 http://bit.ly/htv3J7 Sean Mooney is up.
Sat Jul 16 08:05:49 2011 Christine Orengo on her team’s work at #AFPCAFA11
Sat Jul 16 08:06:28 2011 Orengo says that they are not really function predictors, but evolutionary classifiers #AFPCAFA11 #ISMB
Sat Jul 16 09:18:03 2011 Shaneka Simmon from Jackson State on predicting functions associated with biofeuls – universal stress protein domains. #AFPCAFA11 #ISMB
Sat Jul 16 09:21:30 2011 Simmons: look for diversity of universal stress response genes. in Rhodopseudomonas palustris. #AFPCAFA11 #ISMB11
Sat Jul 16 11:33:52 2011 http://yfrog.com/kevdmbqj #AFPCAFA11 #ismb discussion panel now
Sat Jul 16 14:46:07 2011 Wyatt shows that paralogs actually give better annotation transfer than orthologs. http://bit.ly/nWShuB #AFPCAFA11 #ISMB
Sat Jul 16 14:47:17 2011 Wyatt’s claim runs contrary to common wisdom. Which is good http://bit.ly/nWShuB #AFPCAFA11 #ISMB
Thu Jul 21 22:42:34 2011 After all the talk about standards at #AFPCAFA11 #ISMB this is very timely: http://xkcd.com/927/

I have recently been invited to Google+, (thanks Ruchira!)  and as social timesucks go, it is quite enjoyable.  Here are my first impressions about g+. In a nutshell: g+’s attempt at compartmentalized communication is laudable, but falls short of the mark. Mainly because, no social network can emulate the real-life way we communicate (or not) with the people we know. Not that g+ ‘s goal  emulating real-life social interactions (for those of us who still have them), but let me explain myself.

Compartmentalized communication is what we do every day: my wife reminds me about my medical appointment, but not in front of the kids, because the little one still freaks out over doctors. At work I tell someone I need to cover for me teaching (because I need to take a day off) but not my other colleagues (none of their damn business). While waiting at the Doctor’s office I receive an email that my paper is finally published. I email the info to all the co-authors, and I may also just put it up on Facebook, because I am kinda proud. But of my 100 or so FB “friends” maybe 30 will understand why this is important to me.

The average Facebook user just likes to post stuff on his wall, assuming interested friends will be reading and posting back. Most FB friends are not real friends, in the social sense, but many are acquaintances, professional colleagues, and people you just do not want to turn away when you get a FB Friend Request.  G+ breaks those myriad people into circles: family, friends, acquaintances, followers. I also added “bioinformatics” and “science”.  This compartmentalization by circles initially makes sense because it is more in harmony with what our social life is supposedly like. We have our family, out really good friends, our drinking buddies, our professional colleagues, people we play sports with on weekends, etc. This is where the “too complex” part comes in: in FB, the network rigidly bins my connections into family, friends, friends of friends and everybody else. These are essentially fixed sets, concentric, that are more driven by privacy than anything else. In g+, the circles are supposed to be driven by interest rather than privacy. Stuff I’d like to share with my professional colleagues goes to the “bioinformatics” circle and/or the “science” circle. Personal stuff goes to “friends” or “family”. Like I said, this makes sense initially, because it follows our ideas of social interactions. But keeping people binned in your head, deciding who belongs where and who is privy to what information form which aspect of your life is something really fuzzy and complex. We do it every day, hopefully well, because we have been developing the social skills needed for compartmentalized communications since we were born. Just think back to the “paper published” situation. You want to inform your professional colleagues, family, maybe some closer friends, but your weekend running mates couldn’t care less. Except those two that you run with and are also scientists. Using FB you would just broadcast this information to all FB friends, and let those who care write comments on your post or wall. Supposedly g+ gives you a better option: post only to scientists, to friends, and family. But what about those two runners/scientists? OK, they may be also in your scientists circle but if they are not, that misses out. Compartmentalizing acquaintances in real life for information is complex enough, and I think that the g+ attempt to emulate this will not work well. This is where the “too simple” part comes in: to emulate your real-life compartmentlization, you need to perform dynamically complex intersections and deletions between those circles. Of course, doing this in real life is complex enough, and few, if any, would care to be bothered tweaking this on a social network for each and every post. Google+ circles is somewhat like the compartmentalization solution offered by FriendFeed rooms, but with a better interface.

What I foresee is a kind of facebook/friendfeed mashup. At most the average user will have handful of circles, with many messages going out to all where precision in targeting messages is sacrificed to ensure that all will get the message.

Oh, and nobody will give up Facebook quite yet, as inadequate as it is. So g+ should implement it’s FB and Twitter import/export plugins to draw in the crowds.

.

3quarksdaily are running their annual competition for best science blog entry. There are 87 really great blog posts competing there for the votes of  3QD readers. But who has the time to read through all of them? So let me make it easy for you: vote for me.

This blog’s “Dead Salmon” blog post has been nominated. To get the prize, I need your help. Only the top 20 entries out of 87 nominees will be considered for the final prizes. Please vote for Byte Size Biology: but did you check your results using a dead salmon? If you vote for me, you don’t waste time going through all the other posts, you know you voted for a great post (read it, it’s totally cool. Seriously). And you get to do you civic duty. Or something.

So vote. Then have your family members vote. Your work mates. Your lovers. Your pets. Your car mechanic, your dentist, your hairdresser, your haberdasher, your dog walker, your parole officer, your tattoo artist, your chauffeur and your bartender. Spread the word. Retweet, and share!

There are some interesting developments regarding the February outbreak of Legionelliosis which was traced to the Playboy mansion. Reminder: over 120 delegates of the DOMAINFest in Santa Monica, California came down with symptoms of a respiratory illness. The convention included a trip to the Playboy mansion, which later was suspected as the outbreak source. The convention was held Feb 1-4, 2011. The first inquiry to the  LA County Department of Health (LAC DPH) of a suspected legionellesis outbreak was made by the media on February 11. When tracing the outbreak, LAC DPH and CDC scientists discovered a trail of reports preceding February 11 in social media,  including Facebook and Twitter. The Wikipedia entry was updated almost at the same time when the LAC DPH inquiry was made.

Source: LA Times. From a presentation by Dr. Caitlin Reed

Note that there is quite a bit of information in this entry already!

The scientists identified several risk factors: staying at either one of the hotels, and two parties. They assessed the exposure risk based on the fraction of people who were in those venues and who became sick. To do so, they circulated questionnaires, via email and a Facebook list. They came up with this:

Source: LA Times. From a presentation by Dr. Caitlin Reed

The February 3 “Venue A” party seemed to hold the largest relative risk (RR), by far.  That was the Playboy mansion. Indeed, checking the water at the grotto where the party was held:

Source: LA Times. From a presentation by Dr. Caitlin Reed

The scientists concluded that use of social media to trace outbreaks has its pros and cons. The biggest pro was the ability to contact all conference attendees quickly, even though they are geographically dispersed. Also, the Facebook list, tweets and email encouraged a fast response to the survey. They did have some red herrings, and had to put effort into rumor control. I am not sure how much the last two were amplified by the social media effect.

The full slide presentation, and more information are at the Los Angeles Times.

Why yes, one of my posts did make it in. However did you guess?