Why more Genetic Chemists should study Computer Science

[MrHacks]

(NOTE: I wasn't sure if I should have posted this in the Chemistry board or the Computer Science board. But I guess since the question is more CS than Chem, I posted it where I thought it would be the most appropriate.)

It's been a very productive week for me this week. Aside from my usual weekly job search (doing it daily will drive you nuts!), I felt like being my nerdy self and learn what all the skeletal diagrams in organic chemistry mean. While I haven't studied chemistry since high school, I do have a little bit of understanding of how chemistry works. (P1*V1/T1 = P2*V2/T2, and some of the basic chemical reactions.) The more detailed stuff (electron orbits, etc.) I could care less about, after all my forte is Computer Science not Chemistry.

So, when I finally understood what all those hydrocarbon chains and pentagons and hexagons mean, I decided to wander about the internet and learn little bit more about the chemicals that make up our DNA and how the sequence of nucleic acids that formed the codons would be translated into amino acids in the mRNA process.

It was at this point my Computer Science spider-senses kicked in and though how this process would be translated into a form of finite state automata.

Consider the input symbols Adenine (A), Cytosine (C), Guanine (G), Thymine (T) (in DNA), and Uracil (U) (in RNA). And consider the 64 sequence of the of these symbols (where T and U are interchangable depending if you are playing around with DNA or RNA, respectively) that produce the 20 amino acids (Google "Codon Table" for the list) and three final states (UAA, UAG, and UGA). With the inital state being the amino acid Methionine (Met or M for short), triggered by the sequence AUG, what would the finite state machine look like?

[mr-mitch]

I'm under the impression DNA is more like a turing machine rather than a finite state machine. You have a tape (mRNA), and your machine tRNA binds to certain sections (input) and results in a certain output (aminoacids). DNA can be somewhat rewritten/mutate when it's put back together.

But yeah, that's about the limit of my knowledge.

[mfb]

You want to use the 64 triplets as states? In that way, I think they are meaningless. You could read individual genes as some sort of (not turing-complete) programming code, where each triplet produces one amino acid as ouput when interpreted by the cell.

There is no general rule which amino acid follows another one in which way. There may be patterns that occur more often than others, related to protein folding, but that is beyond my knowledge.

[WarDaft]

I'm under the impression DNA is more like a turing machine rather than a finite state machine. You have a tape (mRNA), and your machine tRNA binds to certain sections (input) and results in a certain output (aminoacids). DNA can be somewhat rewritten/mutate when it's put back together.

But yeah, that's about the limit of my knowledge.

Nothing in the world is actually more like a Turing Machine than a FSA. Computers only approximate Turing Machines by being FSAs with very large memory.

[dissonant]

http://en.wikipedia.org/wiki/Genetic_code is well worth a read. I know some genetic chemists, and they have no problems with computer science.

[Xanthir]

Nothing in the world is actually more like a Turing Machine than a FSA. Computers only approximate Turing Machines by being FSAs with very large memory.

That's not *quite* accurate. Really, physical computers are closest to Linear Bounded Automata, iirc. (LBAs are TMs, but the tape is limited to a linear multiple of the input size).

[WarDaft]

But any given computer is limited to just a fixed size. Doesn't that make them FSAs with a massive state pool?

[korona]

Yes it does but the TM concept describes them in a more useful way. The question is whether this model gives you an interesting theory; the FSA does not.

[markop2003]

"And consider the 64 sequence of the of these symbols"
Why?
64 triplets means nothing in biology...

A finite state machine wouldn't really be of much use for explaining Translation at the level I covered it it's kinda like shaking a box of jigsaw pieces and eventually they all fit together, you'ld just end up with a different state for every single amino acid which dosn't really provide any benefits.

[Karantalsis]

I'm not a genetic chemist, I am a molecular biologist (or genetecist if you prefer). There are so many things that you are failing to account for, such as codon usage by organism, translational codon recognition wobble and mispriming (to name but a few) that understanding more of your discipline, to the point where I could follow what you mean by the question of what the finite state machine would look like begins to look pointless immediately, as whatever it looks like, it will not be very similar to the life it is trying to model at all.

There are alot of people who make in silico simulations of biology, attempting to use computers to model various behaviours, but whilst they are getting better they are still not very good and can only model highly constrained situations. There are one or two good models in existence, but only of a few very specific things. The field is in its infancy still, but if you are interested in the idea of computers modelling biology I'd suggest looking up the in silico heart model, as well as integrative biology, systems biology and bioinformatic techniques.

[Izawwlgood]

Sigh.