I am reading the paper "Neutral evolution of mutational robustness". Because I am new to neutral evolution, I have a few questions.

  1. Firstly, from my understanding, genotypic neutrality means the genotype is not responsible for an increase in fitness. What else can we say about genotypic neutrality? Does it have something to do with the population being randomly selected?
  2. Secondly, how to understand a neutral network exactly? Another article says "by linking every pair of genotypes that are mutually accessible through mutations, genotypes organize themselves into neutral networks", so what means "mutually accessible through mutations"?
  3. Thirdly, the neutral network is like an assumed black box we know nothing about, is it right? That is why in the original article with title "neutral evolution of mutational robustness", the authors mentioned that "one can infer important structural information about neutral networks from data on evolving populations."
  • 1
    $\begingroup$ Thank you for editing @AliceD. I didn't even tried to read the post the first time! $\endgroup$
    – Remi.b
    Commented Jun 17, 2015 at 3:04
  • 2
    $\begingroup$ Can you add a link to the two papers and indicate where the citation come from in the papers? I never heard the term of genotypic neutrality but saying the genotype is not responsible for an increase in fitness sounds like there is some semantic issue in your mind. I don't know either the concept of neutral networks. So the papers would help a lot. You should always restrict your post to only one question. $\endgroup$
    – Remi.b
    Commented Jun 17, 2015 at 3:08
  • $\begingroup$ I'd suggest to divide the current post into two and ask for explanation about the meaning of one particular sentence. $\endgroup$
    – Remi.b
    Commented Jun 17, 2015 at 3:08

1 Answer 1


Mutations in DNA or RNA sequences do not necessarily result in significant changes in the functions of the proteins they encode (or in the case of RNA ribozymes, ribozyme function ). This is because, for a variety of reasons, the change in DNA/RNA sequence may not significantly alter the structure and function of the ribozyme/protein (the function that contributes to the fitness of the organism). If two alleles of a gene with different sequence have equivalent function (again with regard to the function of the gene that provides fitness) the differences between them are selectively neutral.

The neutral network consists of every variation in the sequence of the gene that can be reached from another sequence (of equivalent fitness) by a single base pair mutation (these aren't the only type of mutation, possible, but they are common and more likely to be neutral than more drastic changes). So for example (using an unrealistically short sequence), if I had a gene with the sequence 0 AACAATGCTGACTGA, and alleles 1 AACAATGCTGACTAA, 2 AACAATGCTGACAGA, 3 AACAATGCTGACTGG with equivalent function, 1, 2, and 3 would be directly linked to 0 in the neutral network as they all differ from 0 by a single base pair. The sequence 4 GACAATGCTGACTAA if of equivalent function would also be in the network as it differs by one base pair from sequence 1.

Each gene in the network being mutually accessible through mutations means that I could start out with any of the above sequences and reach any of the others through a series of single base pair mutations that traverse the network (i.e. I could mutate 0 into 4 by first mutating to 1, another network member, then to 4).

The genotypic space includes the sequences of all extant alleles of the gene. This paper suggests that there is a tendency for these sequences to form these networks and especially to evolve towards highly interconnected networks. This is the structural information the author refers to. It should be easy to imagine how this might confer robustness to deleterious mutation. Let's think about our sequences from above, imagine if every sequence was only linked to two other selectively neutral sequences forming a long chain. The sequence AACAATGCTGACTGA is 15 base pairs long with 45 possible single mutations, if only two of those are selectively neutral, the chance of mutating to a neutral network member is low, and even if you do hit on a neutral mutation, your chance of hitting another neutral mutation next time (reverting back or moving on down the chain) is still only 2/45. Now imagine that the neutral network is highly interconnected, with sequence 0 linked to sequences 1, 2, 3 and several other neutral 1 base pair changes, imagine those sequences are linked through each other by other neutral intermediates (being 2 base pair mutations away from 0). Not only is the chance of a single mutation being neutral higher, but subsequent mutations (remember that these include reversions from prior mutations) are more likely to stay within the network (you're more likely to come back around to 0.

Would you have more long-term success walking around a fitness plateau or walking across a fitness rope suspended over the abyss?

  • $\begingroup$ Oh this is what a neutral network is +1! I never heard of this term before. The concept is closely related to the concepts of adaptive landscape and quasi-species. Can you please add a reference that uses the term neutral network as you defined it? Is this term most often used in the genetic algorithm literature? You talk about the authors. I am not sure which authors you are talking about. $\endgroup$
    – Remi.b
    Commented Jun 17, 2015 at 21:41
  • $\begingroup$ Ah, the original question before the edit sounded as if it was referencing this paper or related literature: pnas.org/content/96/17/9716.full.pdf I don't know the body of the literature well enough to comment on how broadly the term is used. I would $\endgroup$ Commented Jun 17, 2015 at 21:49
  • $\begingroup$ Thnks. I edited the post to add the reference. Feel free to roll back and place the reference elsewhere if you prefer. $\endgroup$
    – Remi.b
    Commented Jun 17, 2015 at 21:54
  • $\begingroup$ Great explanation. Two questions. (1): In the last sentence of the forth paragraph, you said "not only is the chance of a single mutation being neutral higher, but subsequent mutations...", I can not see the reason that the chance of a single mutation being neutral get higher. (2): In your last paragraph, long-term success means being more firness? This is an interesting question. $\endgroup$ Commented Jun 18, 2015 at 14:01
  • $\begingroup$ (1) At any given node in the network (a given genotype) the chance of the next mutational event being neutral is higher if that node is connected to many others (assuming the event is random the chance of a neutral event is (possible neutral mutations)/(all possible mutations)). If the new node also has many connections, the chance of a neutral mutation is likewise higher, and so on. Therefore a network in which most nodes have many connections to other nodes (a highly interconnected network) will be more robust to mutation than a network in which most nodes are not linked to many other nodes. $\endgroup$ Commented Jun 18, 2015 at 14:18

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .