Computer independently solves 120 year old biological mystery

[n2doc]

For the first time ever a computer has managed to develop a new scientific theory using only its artificial intelligence, and with no help from human beings.

Computer scientists and biologists from Tufts University programmed the computer so that it was able to develop a theory independently when it was faced with a scientific problem. The problem they chose was one that has been puzzling biologists for 120 years. The genes of sliced-up flatworms are capable of regenerating in order to form new organisms -- this is a long-documented phenomenon, but scientists have been mystified for years over exactly what happens to the cells to make this possible.

By presenting the computer with this problem, however, it was able to reverse engineer a solution that could explain the mechanism of the process, known as planaria. The details discovered by the computer have been published in the journal PLOS Computational Biology, along with the artificial intelligence method used to develop the theory.

The significant thing that the two researchers Daniel Lobo and Michael Levin were hoping to discover was not how new tissue is generated, but how it knows what shape and proportions to grow in. That information is locked away in our genes.


more

http://www.wired.co.uk/news/archive/2015-06/05/computer-develops-scientific-theory-independently

[bananas]

1. it's not a scientific theory, just an untested hypothesis

As pointed out on another website:

http://www.popularmechanics.com/science/a15886/computer-scientific-theory/?fb_comment_id=fbc_907200456005842_907651815960706_907651815960706#f156464b427ba37http://www.popularmechanics.com/science/a15886/computer-scientific-theory/?fb_comment_id=fbc_907200456005842_907651815960706_907651815960706#f1564

Jeff Luckett · Auburndale, Massachusetts

Not a theory ... a hypothesis.

Doesn't get to become a theory until it's been thoroughly tested.

Reply · · 7 · June 5 at 9:34am
[hr]
Arlo Kleijweg

It *was* thoroughly tested. Just because the tests were done before instead of after doesn't invalidate the theory.

Reply · · June 5 at 1:10pm
[hr]
Matthew B. Kilgore · St. Louis

Arlo Kleijweg Actually it does. Unless this hypothesis was validated by a subset of the data that was not used in the creation of the original hypothesis. I do not see this article mention such a test. Without such a test it is just data fitting.

Reply · · 8 · June 5 at 5:34pm

[bananas]

2. how it was done

http://www.popularmechanics.com/science/a15886/computer-scientific-theory/

Simply put, Levin and Lobo's computer attempted to mimic real-life studies over and over again in an excruciatingly-detailed simulation. The machine would randomly guess how the worm's genes formed a regulatory network that allowed for this amazing regeneration, then let that genetic network take control in a simulation, and finally measure how close the results were to real experimental data. If its guesses were good (meaning the gene network made the simulated worm regenerate similarly to real-life experiments), then the machine slightly modified the random genetic network it had created, and tried again until its model was even better.

This may sound simplistic, but after three days of constant guessing, simulating, evaluating its guesses, and tweaking its tactics, the computer had invented a core genetic network that faithfully matched every one of the hundreds of experiments in its database. Essentially, it had explained what scientists had failed to—how the genes connect.

<snip>

Although the computer needed just three days to solve the flatworm genetics problem, Levin says that it took years and years to design and prepare the computer program. For one thing, the duo had to track down hundreds of scientific experiments performed on flatworms just to translate those experiments into a massive database for the computer—essentially giving the machine rigidly structured raw material. They even had to devise a personalized formal computer language that fit the data they needed to describe.