Gene2Music

May 17, 2007 9:17 PM

Gene2Music: "We assigned a chord to each amino acid," said Rie Takahashi, a UCLA research assistant and an award-winning, classically trained piano player. "We want to see if we can hear patterns within the music, as opposed to looking at the letters of an amino acid or protein sequence. We can listen to a protein, as opposed to just looking at it."
posted by Alvy Ampersand (30 comments total)

I don't really see how assigning fairly arbitrary tones to this kind of data would reveal anything interesting about it. It seems like this type of encoding of the data would be more difficult to search and find patterns in than a visual encoding.
posted by demiurge at 9:25 PM on May 17, 2007

Somebody did this when I was a grad student - my friends who heard the result said it sounded just like noise.
posted by pombe at 9:28 PM on May 17, 2007

"The result was a short burst of the most hideous cacophony in G minor. So much for random shortcuts."
posted by weston at 9:36 PM on May 17, 2007

Well, if you look at gene sequences in short, they look pretty much like noise. See if you can tell which one of these I copied from the human genome project and which I generated from random numbers:

1. gaggtaagtggctgtttgctgaggctagtcttcattcacatgtg
2. ctttcctaaaaatgcaacagataatgctgctagattgttattttg
3. tctcctttaacccgccgaacaacctacagcggtaatagatattg
4. gattggcattttaaaatcggtatttaaactgaagacattgtcat

I understand that these audio clips were generated from amino acid sequences, and not base pairs directly, but without some kind of data showing that it actually helps understanding, I'm not sure it gives any insights into music or biology.
posted by demiurge at 9:53 PM on May 17, 2007

It gives insights into the nerdy* intersection of music and biology.

*not that there's anything wrong with that
posted by b1tr0t at 9:55 PM on May 17, 2007

If there's any perceptible pattern, it's going to be in waltz-time...
posted by Steven C. Den Beste at 10:16 PM on May 17, 2007

I can hear the dreadful skirling of bagpipes coming through my open apartment window. This is unusual, since I am Vancouver, and not, say, Fife. The tunes they are playing - for such i must call them, in lieu of a more immediately suitable word - indicate quite strongly that they, too, have stumbled across the same web page as I.
posted by Jon Mitchell at 10:21 PM on May 17, 2007

Just auralizing a data stream without consideration of human pattern-detecting abilities isn't going to do much.
posted by anthill at 10:38 PM on May 17, 2007

If there's any perceptible pattern, it's going to be in waltz-time...

Why? They are not sequencing nucleotide codons, they are sequencing amino acids.

To musically differentiate one protein from another, it would have been more useful to somehow encode secondary or tertiary structure into sound.

Structure is, functionally, how proteins differ. It just takes a critical residue change and the protein can fold into something with different characteristics.

This sequencing algorithm would not seem to allow the human ear to distinguish between two structures efficiently.
posted by Blazecock Pileon at 11:26 PM on May 17, 2007

"... the most hideous cacophony in G minor."

Yeah, cacaphony is so much better in D minor. The saddest of all keys.
posted by flapjax at midnite at 11:31 PM on May 17, 2007

My EEG reader software will "play" brain waves for me. You can definitely hear it when a seizure causes abnormal synchronization of the rhythms.

But it's extremely annoying. I prefer to do it with my eye only.
posted by ikkyu2 at 11:43 PM on May 17, 2007

I can only think this is some kind of pandering/condescension to a non-science population -- an analogy they think even the common people will be able to grasp or something. Because what biologist would want to do such a meaningless dumb thing?

Gene sequences may look like noise (and are, unless you know the reading frame, granted) but an amino acid sequence is a pattern -- a pattern a molecular biologist can read like a sentence. (And a computational chemist can read like a manual.) This guy's arbitrary chord assignments obfuscate the pattern that is there, replacing it with noise.

I can do it, too! Here, drink this -- it's the front page of Metafilter in flavored syrup. Now I will play Beethoven's fifth symphony in scented oils. Next I will slap this dodo bird and take his job.
posted by Methylviolet at 11:50 PM on May 17, 2007

Someone also played a tune with urine samples. Then there was some other biological matter put to music. I don't remember what it was but I'm pretty sure it sounded like shit. What's next, musical toenail fungus?
posted by IronLizard at 12:40 AM on May 18, 2007

What's next, musical toenail fungus?

Toenail fungus had a couple of tunes on the British charts a few years back, but the third single went nowhere and their label dropped them. They've still got a bit of a following in Japan, though.
posted by flapjax at midnite at 12:46 AM on May 18, 2007 [1 favorite]

I think they changed their name to huggabroomstik.
posted by IronLizard at 12:54 AM on May 18, 2007

The "music" is surprisingly engaging. It doesn't sound random or non musical really. Since each "note" is actually a three note cord it doesn't sound as weird as if it were all single notes being banged out on the piano. And I guess it's all major chords so there's never any dissonance or what ever the proper term is for that is. Just a sort of insane piano player. I dunno, someone who knows music theory well could probably explain it better.
posted by gigbutt at 3:47 AM on May 18, 2007

The Shamen, a british techno/electronica group, did this on their track "S2 translation" (sample here) from 1995. I liked that album :)

Seems like there's a lot of prior art (including the Shamen) mentioned on this page: Genetic Music Sourcepage.
posted by reptile at 4:38 AM on May 18, 2007 [1 favorite]

See if you can tell which one of these I copied from the human genome project and which I generated from random numbers:

1. gaggtaagtggctgtttgctgaggctagtcttcattcacatgtg
2. ctttcctaaaaatgcaacagataatgctgctagattgttattttg
3. tctcctttaacccgccgaacaacctacagcggtaatagatattg
4. gattggcattttaaaatcggtatttaaactgaagacattgtcat

The second and the fourth one were copied.
posted by grouse at 6:06 AM on May 18, 2007 [1 favorite]

Someone also played a tune with urine samples.

I heard something a while back about a system that played bits of classical music, but modulated the notes based on patient lab results. The further from ideal the clinical results the more not-right the note got. If Ode to Joy comes out, "Duh duh de de duh Dwooooooterrrrzzzzzzz duh duh...." maybe you might want to go back and give those test results another look.
posted by Kid Charlemagne at 6:16 AM on May 18, 2007

demiurge's fourth sample ends with 'cat'. Could this be the beginning of lolgattggcattttaaaatcggtatttaaactgaagacattgtcats?
posted by lukemeister at 6:47 AM on May 18, 2007 [1 favorite]

That's going to be hell to follow on "Guitar Hero."
posted by misha at 7:53 AM on May 18, 2007

ELVISLIVES
posted by gurple at 9:06 AM on May 18, 2007

(er, for non-protein-geeks, ELVISLIVES is a non-naturally-occurring but legal peptide sequence)
posted by gurple at 9:08 AM on May 18, 2007

Very good, grouse, and I see it says on your profile page "PhD Student, Computational Evolutionary Biology".

My point was more that non-experts would have no clue. I made them, and I had no clue. Did you run them through a program or eye them to figure it out? Do you think putting them in a musical form would help you out at all in your identification?
posted by demiurge at 9:12 AM on May 18, 2007

My point was more that non-experts would have no clue.

I know, I was just being goofy. I used
SSAHA, and it told me which sequences were in the genome almost instantaneously.

Do you think putting them in a musical form would help you out at all in your identification?

Well, if they were long enough and had an extreme G+C content, then maybe. But if it were extreme enough to tell then I'd probably be able to see by eye ("wow, this sequence is made up almost entirely of As and Ts!"). I agree that this whole thing is kinda silly, and what's more, I've seen similar ideas before at least twice.
posted by grouse at 9:29 AM on May 18, 2007

Just for fun, I thought I would search protein databases for the closest match to METAFILTER:

>ref|YP_001208313.1| diaminopimelate decarboxylase [Bradyrhizobium sp. ORS278]
emb|CAL80098.1| diaminopimelate decarboxylase [Bradyrhizobium sp. ORS278]
Length=446

Score = 29.1 bits (61), Expect = 30
Identities = 9/10 (90%), Positives = 9/10 (90%), Gaps = 1/10 (10%)

Query 1   METAF-ILTE 9
          METAF ILTE
Sbjct 235 METAFRILTE 244

In case you were wondering, that's not a very good match.
posted by grouse at 9:41 AM on May 18, 2007

and the real amazing thing is - if you then assign different chords to each amino acid, the proteins sound different than at first!

science is AMAAAAAAZING!
posted by quonsar at 10:08 AM on May 18, 2007

if we assigned pheremones to IP octets we could smell the web.
posted by quonsar at 10:11 AM on May 18, 2007 [2 favorites]

quosar deploys the elusive and devestating DOUBLE SNARK.
posted by heresiarch at 1:01 PM on May 18, 2007

Made me think of Ulam Spiral and Sacks Number Spiral for some reason.
posted by junesix at 3:27 PM on May 18, 2007