Albert Cipriani

I hate to be a party-pooper (all right, I lied), but in that article’s microscopic print buried as a footnote on the last page I read the following: How realistic is that!? According to Talk Origin’s Chris Colby here:
I majored in English, not mathematics. Still, those are mighty big numbers. I interpret Chris to mean that a real gene only mutates once every 100 billion replications. But the virtual Avida genes for purposes of this study were programmed to mutate once every 4 replications. Isn't that stacking the deck? It doesn't seem reasonable to me. – Sincerely Albert the Traditional Catholic

{Rufus: edited for number clean up}

RufusAtticus

Colby's numbers are too high. The typical per nucleotide mutation rate is between 1e-9 and 1e-10 per cell division. Yes the experiment's mutation rate was larger, but the population sizes were smaller, and they couldn't wait thousands or millions of years like nature can.

pz

Yeah, and that's per nucleotide. When considering the rate per gene, we're up around 1e-6.

I don't quite see how this is a valid objection to the paper.

Albert Cipriani

Now Wait A Minute (times 100 billion minutes!) Rufus,
You can’t just fillip a zero off 100 billion as if it were a gnat and call it a day. You guys are celebrating the proof of evolution here based upon virtual mutations that deviate by many many orders of magnitude from the biological facts.

I don’t understand your use of the word “but.” I thought genetic drift was exacerbated by smaller populations. Ergo, the Avida’s skewered results due to the artificially high mutation rates must have been all the more skewered due to the smaller population.

You say the Avida folks were justified in stacking the deck to get their results because: Come come. The fastest replicating bacteria still can’t do it in less than say 20 minutes, right? The Avida virtual life forms, being replicated by a bank of 64 Pentium chips, have to be doing it millions of times per second. If that’s true, we’re talking about 2,000 million virtual replications per one real replication.

If we give evolution 1 billion years to have done its magic on this earth, that gives the fastest replicating bacteria (every 20 minutes) 26,280,000,000,000 replications. At a million virtual replications per second, the Avida program could run that many replications in just 304 days, not “thousands or millions of years” as you say. – Sincerely, Albert the Traditional Catholic

Arikay

The thing though is what I see here is that people are celebrating not because it proves evolution, but because it shows that Ireducibly complexity can come from a non inteligent designer (unless the computer is now inteligent. )

At least, thats what I get out of it.

Nic Tamzek

The only things that I see the mutation rate affecting are the cases where they had a detrimental mutation that was "rescued" by a later mutation such that the two had a net benefit.

However, such things are known to happen even in real life with real mutation rates (I think), so even this phenomenon is not an artifact of the mutation rate.

nic

RufusAtticus

Yes drift is exacerbated in smaller populations. However, 3,600 individuals although small doesn't generate much drift on the short term. Besides the mutation rate cancels out the main result of drift: loss of genetic variability.

Stacking the deck?

And you know this how? I don't remember any calculations in the paper about how long a generation took.

You need to remember that they're not dealing with a replicator with billions of individuals mutatable parts. Their replicators only had fifty or so instructions. If they used a nucleotide mutation rate, then the probability of having getting a mutation in generation would be very low. However, in real life we know that the proability of having a point mutation in a generation is ~100%. You and I inherited anywhere from 90-900 point mutations from our parents. So, in fact, scaling the per op-code mutation rate to the size of the individuals and the population is a very reasonable simulation of real life. Otherwise they'd be dealing with invariant a-life.

While you're at it, why don't you complain that they rewarded EQU and NAND functions, since nature doesn't reward such thing?

Salmon of Doubt

Why does the mutation rate matter? Surely the only thing affected by the mutation rate is how long it takes to get a result, and not the actual result itself?

Lobstrosity

If mutation rates are too high, it can be detrimental to the population. First of all, it magnifies the number of deleterious mutations in each generation, thereby increasing the number of individuals per generation who may not be fit enough to survive long enough to reproduce. Death rate could exceed birth rate over time and extinction could ensue.

There is also another problem with having mutation rates too high. The mutation rate is a contributing factor to the size of the steps organisms will take through their fitness phase space. This phase space is a highly complex surface marked with countless extrema. Evolution drives individuals towards fitness maxima, but if the steps it takes are too large, it may never be able to settle into these maxima. It may perpetually overshoot* as beneficial mutations are coupled with deleterious ones.

Now, if mutation rates were too low, it seems feasible that organisms might have trouble adapting to rapid catastrophic climactic shifts. I would think those that rely on sexual reproduction have a safety net in that the sexual reproduction maintains via random crossover a high level of genetic diversity within the species even if individual mutation rates are low. This pre-existing diversity might be enough to save the species from extinction at the hands of such an event.

In short, there's a reason why scientists aren't breeding super creatures by bathing animals in glowing green radioactive goop (well, maybe they are in secret government facilities somewhere, but I don't think they are). It's not as simple as "increase mutation rate and increase evolution rate." I believe the relation looks like a monotonically increasing function only up to a some critical mutation rate due to the discrete nature of populations (there are a finite number of organisms in any given population so once mutations per generation exceeds a certain average number per organism, bad things would start happening that ain't helping evolution).

*I could give a example of this by talking about training artificial neural networks, but that might go beyond the scope of this discussion (if anyone's interested, feel free to ask and I'll try to justify my postion with this example).

Salmon of Doubt

Ah, that makes sense, thanks! But does the issue of mutation rates being different invalidate the experiment in the way Albert seems to think it does?