Mad Prime

Was kind of sad this year having a "retreat" in a slightly different building in the same city. I amused myself by constructing a genetics "buzzword bingo" list.

1. Slide showing a signaling pathway with >= 15 proteins named
2. Messing with this gene causes cancer
3. Anything involving stem cells
4. Microarray data
5. High throughput / deep sequencing
6. RNAi
7. Animal model vaguely resembling a human disease
8. GWAS
9. Messing with this gene makes this tissue/organ look funny
10. Epigenetics
11. Mass spec data
12. FACS
13. Evolution
14. A photo that makes you lose your appetite
15. Apoptosis is mentioned
16. HAIRBALL (aka. "interaction network")
17. Anything related to sex (eg. chromosomes)
18. RNA splicing
19. Bacteria
20. Yeast
21. Plant
22. Worm
23. Fly
24. Fish
25. Mammal

Although silly, I found this actually helped keep me paying attention to talks. I applied it to the last session and almost got a BINGO, but Norbert Perrimon's signaling pathway slide only had 13 proteins. So close!

This last Saturday Chris and I participated in a code jam organized by a friend using processing, a programming language oriented towards enabling interactive visualization and graphics. I used processing to create all of the graphs in the recent Nature Biotechnology paper - I came to it because I was frustrated with an inability to create exactly the graph type I wanted. In the end each of my graphs is created by a small program, but I'm happy with putting in that effort to get something that looks great.

For the code jam I had been wanting to create a browsable tree of life akin to the fractal-like tree of life comic I made a while back. To this end I created a Newick format tree, working my way back from humans. I filled in only organisms and names that were familiar to me, reasoning that obscure organisms only serve to make a tree confusing, cluttered, and unapproachable. So far I'm only as far as unikonts, you can get a copy of it here if you like. (Please consider this as licensed under CC-by-SA if you'd like to use it.)

Chris found and applied some java code for interpreting a newick format file and creating a tree data structure, and I worked out a simple recursion for drawing the tree out to a given depth from a given node. Clicking on a node redraws the tree from that location; pressing any key zooms back out by one level. Here is a screenshot of the base:

One of the nice aspects of processing is that visualizations should be portable to java applets. Although we haven't done it yet, hopefully we'll be able to do that. If you have processing installed, you can download a copy of what we wrote. (You'll need to fix the file path in the source code to point at the tree.) Other improvements I'd like to do: to make zooming smooth rather than jumping (this can be disorienting when a large change occurs) and add the ability to search and choose an organism, then color the tree according to distance from that organism so you can click "warmer" colors to get closer to it.

The theme of the event was "clocks" and although I had intended to work on this anyway, it did fit into the theme fairly well: we called the program "Evolutionary Time". We were thrilled to be awarded the "wealth" prize by processing founders Ben Fry and Casey Reas: a gift certificate to the MIT Press Bookstore, which I intend to spend on a reference or two for processing, hopefully to further improve our program. ;-)

You may have seen him before: Snowball the Dancing Cockatoo. ScienceNOW Daily News reports on a Current Biology study - a neurobiologist went to study the bird and sees dancing as a phenomenon connected to the ability to imitate sounds. A companion paper looks for dancing ability in a number of animals, but only found it in animals that were able to vocally mimic sounds.

If you haven't watched the video you definitely have have to. If you have... it's worth watching some again. :-)

The hard copy of the Nature Biotechnology with my article in finally came out, so here's a pic of me posing with it. I designed the cover too, it's an abstract representation of padlock probes capturing sites in bisulfite-treated DNA, with a color-warped photo of induced pluripotent stem cells in the background.

If you happened to stumble across this site because you're interested in the profiling methods, please feel free to contact me. Of course, the news about 5-hydroxymethyl-cytosine kind of throws a wrench into the gears of any DNA methylation profiling studies.

Today I produce things of questionable value to society. Five cheezburgers pleez, kthxbai.

moar funny pictures

Ragdoll cats are remarkably calm. Here is my cat pendulum demonstration of this, starring Zoe. I think she only started to object when I started trying to do the rotation motion.

Today I finally framed a painting of California poppies I bought on etsy a while back. You can click on it to see a more detailed photo of the painting itself. The artist retains copyright on the image, please ask her for a print if you'd like one for yourself! :-)

Yesterday there was a bad smell that kept cropping up in the lab. We thought it might have been the floor drain that is connected to the sewage system; it has a U pipe that should block gas exchange when filled with water but sometimes that evaporates off. So A poured a liter of water into it, but the smell appeared again.

Eventually we discovered it was coming from the 37C incubator where B and C were culturing bacteria as part of a series of experiments I have unflatteringly dubbed "poop-omics". "Well what do you expect if you culture up a bunch of random ass bacteria!" I said.

In this case the Monroe Interpretation regarding the position of "ass" hyphenation would be the correct one.

The other day I was recalling a required course at Caltech, a science communication writing class that is meant to help students learn how to communicate technical ideas. I remember writing for this class (although I can't remember what I wrote about!) and in retrospect I'm saddened to think about how many hours I spent in college writing papers that nobody ever read again. There is a large pool of wasted efforts going into writing papers for college coursework.

It occurred to me that it would be wonderful if this sort of course actually had students choose Wikipedia pages that need attention and completely rewrite them. Writing a scientific/technical wikipedia topic addresses many of the same writing skills that science/engineering students need to learn: the article needs to be accessible as possible while remaining technically accurate, it needs to cover a lot of different pieces of information while maintaining coherence as a whole, and it needs to be well referenced (and, by implication (one hopes), well researched).

It has been pointed out to me that I'm not the first with this idea and that it has been done -- and while they did run into some difficulties it also showed that writing with purpose motivated students to put more effort into their work. I think many of potential pitfalls could be addressed with appropriate foresight. Some concerns are:

• students feeling pressure writing under the spotlight of the public eye
• teachers feeling stressed at having to evaluate whether work is good for Wikipedia rather than focus on helping the student learn to write
• social friction with Wikipedia as students make mistakes in wiki formatting and are unfamiliar with standard protocol

Here is my proposal that I think would address many concerns:

Set up a private mediawiki for the class (or whole school?) into which all papers are written. Students can see each other's papers should they choose, but I don't see that as a problem. Classes require students to show their work to other students all the time, and in this case it's only if they choose to look and they don't have to evaluate each other (unless maybe you want that to be part of the class). Papers are written into this wiki and instructors read them there. You could even have a week or two "free for all" editing period near the end where students can improve each other's articles.

At the end of the course, students can give consent for their work to go onto Wikipedia. It's optional, but they don't have to make the decision until after they're done.

Assign someone a TA position as Wikipedia liason for the group - not a writing expert, instead someone familiar with Wikipedia, like me. (After an iteration or two of this, prior students are likely to be interested in taking this TA position. If that doesn't happen perhaps the course isn't worth continuing.) This person has the following tasks:

1. evaluate the students' initial choices for topics to write on - check that the target article is poor quality and deserves some rewriting work
2. once these are chosen, give notice on the Wikipedia talk pages the possibility of a page rewrite in the next few months.
3. At the end of the course, for those students who give consent to pushing the work to Wikipedia, evaluate whether their versions are an improvement on the article. If there are sections of the article that were lost and should have be preserved, communicate with the students to integrate those into their work. If there are wiki style problems they can tell the students what to correct. Give notice on these pages about imminent rewrite. If there are wiki style problems they can tell the students what to correct.
4. The liason can also communicate with relevant Wikiprojects if they doubt their own qualifications for determining the factual accuracy of the rewrite.
5. Update the pages with the rewrites.

This can be a custom Wikipedia account eg. "MIT Wikipedia Liason".

This way instructors can focus on improving the writing and not worry about evaluating whether it's good enough to integrate. Students won't be pressured to make their work public, it only happens if they decide to do so. The liason makes sure the transition of material into Wikipedia goes smoothly.

Writing a wikipedia article could be an option within a course rather than a requirement (ie. an alternative to standard dead tree paper writing) and this set-up could accommodate multiple instructors and courses should multiple people be interested in offering this style of writing as an option.

One of the more profound discoveries I've made regarding color blindness is that there are only two hues in a color blind person's world: blue and yellow. For some reason I thought that the hues between blue and green were still vivid colors, like this:

But actually it looks like this (deuteranope simulation):

You can see here that turquoise looks the same as gray/white -- in other words, it looks colorless. There are really only two hues: anything between them looks less intense, more gray.

White light is a mixture of all colors - it activates all receptors equally. Because turquoise is right between the two receptors it also activates them equally. The two types of light are providing the same information, it takes a third receptor with a different response to tell the difference. Having a third receptor has a profound effect: all wavelengths become distinct colors, a rainbow of hues is visible. Would adding a fourth receptor have a similarly profound effect? I don't think so -- the spectrum is a linear one-dimensional type of information -- but maybe I'm lacking imagination here.

Green traffic lights actually have a bluish tinge to them to distinguish them from the red and yellow lights. Because of this, they actually look white! Here is a digitally merged photo I took, along with a deuteranope simulation:

This colorless white/gray effect for hues that hit both receptors evenly is also visible on the other side of the color wheel, in the "unnatural" hues formed by mixing red and blue.

Thus the colors potentially confused by red-green color blind fellows goes beyond distinguishing between hues in the red-to-green range. Turquoise and magenta can be confused for gray, and purple can look blue. I'll close here with a series of potential color confusions: