Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Science

Computational Genomics 34

Posted by michael from the throw-more-cpu-at-the-problem dept.
blamanj writes "Scientists at UC Santa Cruz have been using computational techniques to 'reverse engineer' the DNA of extinct species. David Haussler and colleagues created a hypothetical portion of ancestral mammalian DNA and let a computer model simulate the process of evolution. Then they made their algorithm work backward from these descendants, to see if it could recreate the original ancestor."
This discussion has been archived. No new comments can be posted.

Computational Genomics More

Computational Genomics

Comments Filter:

  • Reverse enginering (Score:3, Interesting)

    by Ender_Stonebender ( 60900 ) on Thursday December 02, 2004 @03:31PM (#10977936) Homepage Journal
    Does this seem like "we'll get the original order of a list based on the sorted order and knowing how the sort algorithm took to run" (in otherwords, bound to be so wrong as to be useless)?

    Or is it just me?

    --Ender
    • Almost, but not quite.

      Which would be closer to taking several outputs of a semi-random "black box" function which is reasonably well understood, and trying to determine the common input that generated the various outputs.

      Even that's not a terrific analogy, really, but it's a little closer.
      =Smidge=

      • I would compare it to analyzing languages spoken today to determine how the language they descend from (such as proto-indoeuropean) may once have sounded. While many indoeuropean languages are mutually unintelligible today, they share certain fundamental elements that are best explained by them having been present from the start. It's not an exact science, of course.

    • Re:Reverse enginering (Score:3, Insightful)

      by Lenale ( 792831 )
      It does sound a bit fishy... I just attended a lecture on DNA-focused biophysics the other day, and they were all about "we won't be able to compute it for years, but..." And by the way, as the article said, we're quite a bit behind the rodents in losing bases... let's make babies :)
      • Biophysics isn't quite the right comparison, since biophysicists tend to deal with problems that are computationally hard. (And I mean "hard" in the strict CS-y sense, not in the "I'm having trouble coding this particular function" sense.) Genomics, while there is plenty of core algorithm work to be done, is much computationally much easier -- there are well-known polynomial-time sequence comparison and reconstruction algorithms, for instance.

      • And by the way, as the article said, we're quite a bit behind the rodents in losing bases...

        The article explained the difference in mutation rates by referring to the shorter reproduction rates of rodents. However, as I understand the process of transferring DNA from one generation to the next, mutations may occur whenever a cell splits in two, not only when the animal reproduces. I seem to recall from Sykes' book The Seven Daughters of Eve that the average number of successive cell divisions in the repr

        • Re:Reverse enginering (Score:2, Interesting)

          by daymitch ( 699517 )
          Think about this for a minute. Rat cells don't have to divide any faster than humans for them to have had more cell divisions since our divergence. They reproduce more often so they have more generations per unit time than we. The rats somatic cells are also probably just 20 divisions or so away from the original zygote. Just multiply it out and you have more cell divisions per unit time in the rats.

          • I believe the answer is in line with your explanation, but I still can't really visualize the process well enough to understand it. If the female rat is one year old and the human woman is 30, how come their respective egg cells are both 20 "cell generations" younger than those of their mothers?

            While the entire rat population will experience a higher number of cell divisions (and thus a proportionally higher number of mutations) per unit time due to its size, those mutations will normally end up in differe

        • mutations may occur whenever a cell splits in two, not only when the animal reproduces

          This is true, but don't forget that most of these mutations are completely lost right away; only those that are passed on to future generations by reproduction are able to persist.

          A mutation which starts in the brain but doesn't make its way to the reproductive organs prior to procreation doesn't get passed on to any future generation.

          • A mutation which starts in the brain but doesn't make its way to the reproductive organs prior to procreation doesn't get passed on to any future generation.

            Of course, but this doesn't affect the mutation rate, and thus won't explain the differing mutation rates between rats and humans. A mutation in a rat's brain cell is no more likely to make it to the reproductive organs than a mutation in a human brain cell, in spite of the rat being a lot smaller (it supposedly has fewer of all kinds of cells, not s

      • Fishy, you say? I immediately thought of hollywood science.

        "Detective, we have a new computer program that can predict the path of any bullet...."
        "Yeah, so?"
        "So we tried running it backwards, and we just found out where the suspect bought the ammo!"

  • 98 percent? (Score:3, Interesting)

    by kureido ( 830125 ) on Thursday December 02, 2004 @03:45PM (#10978104)
    From the article: "Then they made their algorithm work backward from these descendants, to see if it could recreate the original ancestor. The ancestor the algorithm came up with had a sequence that was 98% accurate..."

    Human and chimpanzee DNA are about 98% similar, too. In that context, 98% similarity doesn't seem that impressive. Maybe someone needs to invent a new benchmark for sequence comparison for species that are already similar?

    • Re:98 percent? (Score:2, Insightful)

      by Anders Andersson ( 863 )

      Since the accuracy with which the artificial genome was recreated in the simulation isn't compared with that of other methods for doing the same thing, the 98% figure doesn't tell us much. For all I know, that could be the accuracy you would get using any method (but I suppose the scientists actually have more simulation data than was presented in the article).

      Likewise, comparing that number to the degree of genome similarity between humans and chimps isn't very meaningful either. Since the article doesn't

      • Forget the 98% figure. It honestly doesn't mean anything at all. The tests are very accurate, but their results are meaningless. They amount to comparing two strings of binary written with a language and compiler that were lost 3 billion years ago, for a computer we don't understand. Measuring how every base pair matches up to every other base pair is useless. There are pieces of DNA that are far more important than others, like the ones that turn other genes on and off, or the ones that code for impor
    • Forget the 98% figure. It honestly doesn't mean anything at all. The tests are very accurate, but their results are meaningless. They amount to comparing two strings of binary written with a language and compiler that were lost 3 billion years ago, for a computer we don't understand. Measuring how every base pair matches up to every other base pair is useless. There are pieces of DNA that are far more important than others, like the ones that turn other genes on and off, or the ones that code for important
      • So please, stop asking for clarification on the percentage, because it just doesn't matter. All it tells us is that we are closely related, but not how close, or in what way.

        That was exactly my point. Why bother to run expensive computational genetics experiments if the result comes out as "Well, we know it's close, but not how close, or in what way." The scientists might as well give a picture of an animal to a sketch artist and say, "Draw this, except more primordial."

        • The percentage that doesn't matter is the similarity between human and chimp DNA, not the accuracy with which the artificial sequence was reconstructed after simulated mutations. While it's true that there is little if any correlation between DNA sequence similarity and similarity of the resultant physiologies, the simulation was only concerned with the DNA sequences themselves, not their manifestations as living creatures.

          For all we know, the initial DNA sequence used for the simulation may have been enti

  • Apply this to birds and we'll get the gene sequence for raptors and other dinos..

    Who's up for Jurassic Park? Anyone?

    • Jurassic Park (Score:2, Interesting)

      by Anders Andersson ( 863 )

      I don't know how well understood the lineage from dinosaurs to modern birds are, but I suspect you would need the genomes from a few species that are not descended from dinosaurs (say, mammals) as well, for interpolation rather than extrapolation of the dinosaur genome.

      Even if we could recreate dinosaur DNA in this way, I doubt we have the technology to turn that DNA into a live animal, or even do a computer simulation of that process. Is anybody working on an open-source biochemical simulator?

  • Wasn't "Computational Genomics" in the Sid Meier's Alpha Centauri Tech Tree?
  • I read the Nature summary, but no real articles .. anyone know how they do the "working backwards" thing? I would've guessed some sort of Bayesian analysis, just like most people use to come up with phylogenetic trees, but it sounds like there's something more interesting going on here.

  • Algorithm testing. (Score:4, Interesting)

    by Fortran IV ( 737299 ) on Thursday December 02, 2004 @07:24PM (#10980949) Journal
    The process is interesting, but their description of how they tested their algorithm is less than confidence-inspiring.
      1. 1) Manually create a set of hypothetical data.
      1. 2) Run a mathematical algorithm to generate new data.
      1. 3) Run the converse of the algorithm on the generated data.
    If an algorithm is truly reversable then, without the necessary randomization, such a process is likely to generate the original data with 100% accuracy. I'd have felt much better if they'd run two independent algorithms against each other: create descendants with ForwardA() and extract ancestors with BackwardB(), then do the same thing with ForwardB() and BackwardA().

    • I agree that the way this is expressed in the article leads to your interpretation:

      To assess their method, they created a hypothetical portion of ancestral mammalian DNA and let a computer model simulate the process of evolution, to generate sequences for its descendants.

      Then they made their algorithm work backward from these descendants, to see if it could recreate the original ancestor.

      However, I seriously doubt they actually reversed the simulation algoritm. Reading the entire article, it sounds mo

  • Planet seeding (Score:4, Interesting)

    by Associate ( 317603 ) on Thursday December 02, 2004 @07:43PM (#10981127) Homepage
    Once we get things like this under control along with teraforming, we can seed barren planets. We can walk the universe like gods. Probably have to kick the old one's out first.
  • ...Barclay's turned into a spider.

Slashdot Top Deals

Put your Nose to the Grindstone! -- Amalgamated Plastic Surgeons and Toolmakers, Ltd.

Close