Garte: The Meaning of "Random" Mutation

That does seem like a good analogy, but it actually misses the point, I believe. Because the bets are not random, meaning people who place bets are not equally likely to bet on a 2 and a 7, They know the odds, and will bet on 7 whenever possible (and they do). As for mutations, there actually are some connections between what the organism needs, and the mutations that do happen. Josh on another thread agreed that “All these mutational mechanism are biased towards changes that are more likely to increase fitness.”

And as you pointed out above, we also know that expressed genes are more susceptible to mutation, resulting in a higher number of beneficial mutations (also deleterious ones, but they don’t count) in those genes, which can appear as if the organism favors mutation sites that provide a benefit. This is probably the explanation of the results of John Cairns in 1988, later confirmed by Susan Rosenberg and others.

So, like Perry, I am still perplexed by the meaning of random when applied to mutations. Clearly cells dont decide to mutate specific genes in response to stress, but there do seem to be many mechanisms in place (including error prone repair, which I dont think has been mentioned yet) that tend to bias the results of mutations in a positive direction. I dont see how that process could be labelled random, unless the definition being used is different from the statistical one I have in mind.


The analogy holds up beautifully. If I have to bet on a particular type of mutation, I’ll bet on a transition, the most common type.

That’s tantalizingly vague. Care to elaborate?

But if we use “bias” in that sense, all mechanisms are biased toward neutral changes, neither good nor bad.

1 Like

The same applies to the environment with fitness being non-random.

What he meant (I think) is that mutations are biased towards conservative amino acid changes, that is changes towards amino acids that are more like the original amino acid. However, there is no specificity with respect to the environment or the specific genetic locus. It is very much like 7 being more common than 2 or 12 in craps.

As you state, this only increases the mutation rate in genes, and those mutations occur whether they are neutral, beneficial, or detrimental. As an analogy, this is like poor people buying more lottery tickets than rich people. Even though this trend exists the lottery is still random.

A scenario I would consider non-random would be a membrane bound protein that sensed antibiotics and then triggered other enzymes to make a specific mutation in a specific gene to produce antibiotic resistance.


And a scenario I would consider to be random would be the absence of any built in mechanism (such as the genetic code’s optimization that includes more conservative substitutions as a result of mutation) that guides mutational results toward beneficial or neutral. The existence of these mechanisms, while not producing a well determined result, as in your example, to me rule out the idea of pure randomness in mutational consequences. In other words even if all mutations were equally likely (they arent, thus hotspots) all the well known biological features that act to steer toward beneficial mutational effects give at least some degree of non randomness to the overall process. I think the mistake is to label everything either random or non random. We need an intermediate term, since this after all, is biology, where nothing is ever all or none.

1 Like

That is not the definition that biologists are using. For biologists, they are talking about the independence between the specific mutation the organism needs and the processes that cause mutations.

I would also have to conclude that you think craps is not random because it has a built in mechanism to make 7 more likely than 2 or 12. If craps were purely random then 7 would be just as likely as 2 or 12, right? Most physicists think quantum events are random, yet it would seem you would claim they are non-random because certain events are more likely than others. In fact, I doubt there are few, if any, process in nature where all possible outcomes have equal probability.

Notice, however, this has nothing to do with the term “random.” Random mutations can be independent or dependent on fitness. The problem is the metaphysical baggage imported into “random”, ignoring the actual mathematical definition which is metaphysically neutral.

You have a much better handle on the mathematics side of biology, so I would be curious as to what type of model you have in mind. Would a Poisson distribution be a good place to start, as I briefly spoke about in another thread?

1 Like

@sygarte, all that “random” means is that the mutations are not entirely predictable from the modeler’s point of view. They can follow a pattern. They can even be related to fitness, but the would still be “random”, if we don’t know precisely where they will happen ahead of time. They can be deterministic. They can be intelligently guided. The only claim is that the model itself can’t fully predict them.

“Random” is an intermediate term between “chaos” and “determinism” because it can include patterns, or a distribution. In fact, it is hard to envision a random variable that does not have some sort of pattern. A good description of “random” is here:

In probability and statistics, a random variable , random quantity , aleatory variable , or stochastic variable is a variable whose possible values areoutcomes of a random phenomenon.[1] More specifically, a random variable is defined as a function that maps the outcomes of unpredictable processes to numerical quantities (labels), typically real numbers. In this sense, it is a procedure for assigning a numerical quantity to each physical outcome. Contrary to its name, this procedure itself is neither random nor variable. Rather, the underlying process providing the input to this procedure yields random (possibly non-numerical) output that the procedure maps to a real-numbered value.

A random variable’s possible values might represent the possible outcomes of a yet-to-be-performed experiment, or the possible outcomes of a past experiment whose already-existing value is uncertain (for example, due to imprecise measurements or quantum uncertainty). They may also conceptually represent either the results of an “objectively” random process (such as rolling a die) or the “subjective” randomness that results from incomplete knowledge of a quantity. The meaning of the probabilities assigned to the potential values of a random variable is not part of probability theory itself but is instead related to philosophical arguments over the interpretation of probability. The mathematics works the same regardless of the particular interpretation in use.

Notice that even if mutations are not independent of fitness, they would still be modeled as a “random” variable, because we cannot fully predict them a priori. In fact, even targeted edits with CRISPR would be modeled with a “random” variable, because these edits (even though they are engineered) are not entirely predictable. We know there are patterns to mutations too (transitions > transversions), but this also does not make them non-random. It just means that they follow a distribution biased towards transitions. They are still “random.”

The whole question of “independent” of fitness (or not) is a red herring, in my opinion. First, in a precise sense, I’m not sure it is true. Rather, if you look at the experiments from which it is derived (The Luria and Delbruck Experiment), it just means that beneficial mutations (in a particular context) are not induced by the environment, but preexist selective pressure. Even this is merely a rule of thumb that is not strictly true (e.g. look at the immune system).

Of course, there does need to be more mathematically precise language through out biology. EES is not taking us that direction, but is clouding the water even further by investing “random” with additional meanings that are just not valid. It would be easier to just remember the standard definition of “random” in modeling. It includes no metaphysical baggage. It does not require a grand rethink of biology, and might even make for more clear explanations of evolutionary science.

Does that make sense @sygarte?


This also is an overstatement. A boundary case in random variables is a variable that can take on only one value. I.e. they are fully determined. We would still call this a “random” variable, one with a particular sort of distribution, one without any entropy. Yes, the term “random” is that flexible.

The distribution of a fully determined variable is, for the continuous case, a Dirac delta function.

By this definition, everything in biology (including natural selection) is random, unless you can think of a biological process whose outcome is fully predictable. I can’t.


Exactly my point. Not just everything in biology, but just about everything in the real world is “random,” which is why statistics and probability are such powerful modeling techniques. It is also why the claim that “mutations are not random” (often made by EES) grates against computational biologists.

Of course mutations are random. Of course they have patterns too, a distribution. We are investing an immense amount of effort trying to understand these patterns, discovering new ones all the time. None of this means mutations are not “random,” it just means that we are understanding the nuances of their patterns more and more. Even if we perfectly understood the patterns, however, we still would not be able to fully predict mutations, and would be modeling them with a “random” variable.

OK, I think we have gotten to the end of this long discussion between you and Perry. Its all about definitions which for “random” are about as distant from each other as possible. I would just add that if everything can be described by any particular quality (like being random) then the quality loses most of its value as a description for any particular thing.


That’s my view on it as well. Biologists are trying to answer specific questions with specific models, and the definitions serve those purposes. First, are beneficial mutations induced by the environment? No. What should divergence of sequence look like after a speciation event? The unpredictable nature of mutations (i.e. randomness) will produce differences all over the genome which means two populations should accumulate different mutations in the vast majority of cases. At the molecular level, we should also see different adaptations to the same challenges in separate lineages for the vast majority of cases.

In my experience, this is the lens that biologists are viewing mutations through.


If we call mutations “random” in the sense I am describing it, we now have a precise language for describing its distribution. That is why it still has value. While everything might be random, everything does not follow the same distribution. This is where a lot of science is happening, better understanding the factors that effect the distribution of “random” mutations.

Yes that is true. It has also been explained to Perry. I’m not sure what his objections are, because he says he wants me to give him a definition, but I already have done so. I’m not sure what his point is anymore.

I emphasize I’m using the standard definition of “random” in probability and statistics. Perry’s meaning is more of the metaphysical one, perhaps even ontological randomness. That is fine if he wants to use that definition, but it has almost nothing to do with evolutionary science, so I’m not sure what it has to do with me.

1 Like

Beneficial mutations in the immune system are induced by the environment. Are they not?

I wonder if this similar to our arguments about nested clades. What you are saying is largely true, but not precisely or totally true. It would be better to speak with more precision, in my opinion.

And I agree with all of this.

Two questions come to mind.

What natural process would you define as being random using Perry’s definition?

With what we know about the processes that cause mutations, what description would you use and why?

1 Like

If we are talking about hypermutation in the variable regions of antibodies, and/or initial shuffling of V(D)J components during B-cell maturation . . .

Beneficial, neutral, and detrimental mutations are induced by the environment, not just beneficial ones. If you are infected by a bug it doesn’t trigger specific mutations in specific B-cells to produce a high affinity antibody in a deterministic fashion. Rather, a whole host of mutations occur in the antibody gene which can change antigen affinity for the worse or better or not at all.

1 Like

Radioactivity is fully random, as are many quantum processes. In biology and chemistry, the mass action priniciple is basically random, but less so in biological systems than in pure chemical ones, since organisms have developed methods to improve the probability of interactions between reactants (membranes, etc) As for mutations, I would call them paritally random. as well for the same reason. There is a theme here. Living systems tend to favor less than totally random processes. It makes life easiter that way.


Aren’t useless ones induced as well?

1 Like

So glad to see @sygarte chime in on this. I was starting to feel like a crazy person or, gasp, a YEC, for finding Marshall’s understanding of random much more meaningful than his detractors’ understanding.

1 Like