Standard models in population genetics look up at the evolution of few loci which impact fitness. The variance in fitness is determined by the genetic variance and the environmental variance (and the co-variance between environment and genetics). In this question I am interested only about genetic variance and about what percentage of the total (additive or not) genetic variance in fitness do 'n' loci explain.


In general, in natural populations, what percentage of the total genetic variance is explained by the 'n'- most important loci?

Here, by "most important loci" I mean loci which variance explain much of the total genetic variance.

In other words, the subquestions are of the kind:

  • how much of the fitness variance does the most important locus explain?
  • How much of the fitness variance do the 3 most important loci explain?
  • How much of the fitness variance do the 100 most important loci explain?

Of course, the answer depends on the population under consideration. Factors that might influence the answers are for example

  • species
  • population size
  • environment stability

Beside this question, I also welcome some insights concerning how different factors are likely to influence the answer.

  • 1
    $\begingroup$ Do you have at least one example of an empirical study of the type you are looking for ? It may be a helpful starting point for us in finding other ones. $\endgroup$ – Barbara Feb 10 '14 at 13:16
  • $\begingroup$ No, unfortunately I don't know any empirical study that investigates my question. $\endgroup$ – Remi.b Feb 10 '14 at 14:25
  • $\begingroup$ And do you have any study, which measures percentual increase in fitness, regardless of the cause ? $\endgroup$ – Barbara Feb 10 '14 at 14:33
  • 1
    $\begingroup$ I think its exceptionally rare for the most important loci to collectively explain more than a small percentage of the total genetic variance, it's a fundamental problem with QTL - looking for something of large effect when by definition we expect most or all effects to be tiny. I'm looking forward to seeing an answer to this one! It's going on my favorites $\endgroup$ – rg255 Nov 20 '14 at 16:27
  • 1
    $\begingroup$ I don't think there's a great answer to this one, partly because fitness is a tough trait to measure. I think more is known for other traits like height, but even there, if we want accurate estimates of the variance associated with 100 alleles, we're limited to humans at best. $\endgroup$ – Daniel Weissman Sep 18 '16 at 17:59

From the statistical point of view, this question is rather vague. One would need a mathematical definition for the term "genetic variance".

In one extreme, if the "genetic variance" merely means the categorial variations of nucleotides (i.e. ACTG) in the pooled genomes of interest, then the distribution of total "genetic variance" vs. loci variation is uniform and only depends on the size of the locus.

In another extreme (among many dimensions of extremes), if the "genetic variance" is only manifest by the organism's immediate "fitness" and only has two values: life and death (on birth), then all the "essential genes" are "the most important" loci. If you're interested in the n most important loci where n > the number of essential genes, then you would first look at the binary genetic interactions in the database such as BioGrid where two non-essential genes would "interact" and change the organism's fitness (in life and death).

Of course none of the two extremes is very interesting in population genetics or evolution, but a statistical question is best phrased by statistical terms. I would try to find the mathematical definition for "fitness variance", too.


For a semi-empirical/informatics study, I think you could start with the simplest organism whose genome is well studied.

  • Choose an organism (e.g. yeast)
  • Assume uniform inheritability
  • Choose a specific measurable phenotype/environment (e.g. the ability to grow on a specific sugar x)
  • Scan each gene in the yeast genome and see its quantitative impact on growth (They're documented in various database)
  • Ignore genetic interaction
  • (Or scan each gene pair/triplet/.../n-cluster to see its impact on growth on x)
  • Try to model your empirical distribution. It's only valid for that specific phenotype/environment
  • Define your "TOTAL genetic variance in fitness" meaningfully and rigorously. "Additivity" would be a very drastic assumption.

My guess as a non-geneticist is that, as rg255 suggested, for each phenotype as a function of the environment, the distribution would follow a power law. They would not have the properties that would allow you to use central limit theorem to "add them up". But for a specific phenotype, your empirical cumulative distribution function (cdf) would answer your question.

  • $\begingroup$ Actually instead of a long edit in my post (now deleted), I will just redirect you to any quantitative genetics book to understand the definition of genetic variance. Consider for example the chapter on quantitative genetics in Gilespie's book. $\endgroup$ – Remi.b Aug 18 '16 at 2:19
  • $\begingroup$ I don't think this answers the question - @Remi.b I think you are looking for something like in species X, the fitness effect of all loci was quantified and ordered, 15% of fitness variation is explained by a single locus, 21% by the top three, and 40% by the top 100, right? $\endgroup$ – rg255 Aug 18 '16 at 7:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.