More Recent Comments

Thursday, November 04, 2010

Human Mutation Rates

Calculating the rate of evolution in terms of nucleotide substitutions seems to give a value so high that many of the mutations must be neutral ones.

Motoo Kimura (1968)
We know a great deal about the error rate of DNA replication. The replisome makes a mistake about once in every 100 million bases incorporated. This is an error rate of 10-8. The repair mechanism fixes 99% of these errors for an overall mutation rate of 10-10. Given the size of the human genome and the number of replications between zygote and germ cells, this translates to approximately 130 mutations per individual per generation [Mutation Rates].

Recently there have been two attempts to verify this calculation. In one, the Y chromosomes of two men separated by 13 generations in a paternal lineage from a common male ancestor were sequenced. The differences correspond to a mutation rate of 0.75 × 10-10 per generation, or almost the same as theory predicts [Human Y Chromosome Mutation Rates]. This is based on the fact that if most mutations are nearly neutral (they are) then the rate of fixation by random genetic drift should be the same as the mutation rate [Random Genetic Drift and Population Size].

The other study, by Roach et al. (2010), compared the genome sequences of two offspring and their parents. By adding up all the differences in the offspring they arrived at an estimate of 70 mutations in the offspring instead of the expected 130 [Direct Measurement of Human Mutation Rate]. This is half the expected value but the study is fraught with potential artifacts and it's best not to make a big deal of this discrepancy.

John Hawks was worried about this last March [A low human mutation rate may throw everything out of whack ] and he's still worried about it today [What is the human mutation rate?].

What John is really interested in isn't the mutation rate per se since we have pretty good handle on that number. What interests him in is Calibrating the Molecular Clock and that's not the same thing. What it boils down to is the number of years per generation—or the number of fixed mutations per million years.

John thinks that if the actual mutation rate is only half of the value we though it was then the dating of many evolutionary events will need to be recalculated. For example, the human-chimp divergence would have to be re-set to eight or nine million years ago. But that's not strictly correct. We don't calibrate the molecular clock by taking the known mutation rate and multiplying by the number of generations then throw in a known value for the number of years per generation.1

None of those values are known for even the most recent events in the primate lineages. What we usually do is work from a fixed point in the fossil record, count the number of differences between species, and estimate a mutation rate per million years. That value is then used to calibrate other divergences.

Sometimes these rates of change can be related to the mutation rate by estimating the generation times and they often seem reasonable when we come up with generation times of,say, 25 years. Even if the known mutation rate were half of the current consensus value, the most reasonable adjustment would not to be recalibrate the time of divergence but to reconsider our assumption abut generation time. Maybe there were twice as many generations per million years.

But this is actually a non-problem right now since the Roach et al. (2010) estimate is not very reliable. I don't think John Hawks should be worried.

1. Plus estimates of the effective population size, Ne.

Roach, J.C., Gustavo Glusman, G., Smit, A.F.A., Huff, C.D., Hubley, R., Shannon, P.T., Rowen, L., Pant, K.P., Goodman, N., Bamshad, M., Shendure, J., Drmanac, R., Jorde, L.B., Hood, L., and Galas, D.J. (2010) Analysis of Genetic Inheritance in a Family Quartet by Whole-Genome Sequencing. Science (Published Online March 10, 2010) [doi: 10.1126/science.1186802]


TwoYaks said...

We usually take fixed points, but there's an increasing trend to authors using uncalibrated clocks with just some vague assumptions about mutation. Particularly in species with poorly documented fossil records. I can think of one pertinent example where an author uses a generic substitution rate to date divergences in some cervids, which happen to line up with climatic events. But what's left unmentioned is that those dates would be sensitive to small changes in the mutation rate.

John Hawks said...

I agree that the Roach et al. (2010) estimate isn't very reliable, which is where I left it last spring. But I think that the Lynch (2010) estimate probably is fairly reliable, and it is essentially the same value.

DK said...

Doesn't uncertainty in effective ancestral population size trump these little two-fold uncertainties in mutation rate and number of generations?

Anonymous said...

It seems common to omit the effects of recombination and non-cell-cycle mutagenesis from the calculation. If we just use copying error/repair rates and the number of divisions in the germ line (presumably averaged between male and female) and generation time (... same again?), we are tacitly assuming that recombinational mutation and interphase attrition can be ignored, and that the only significant source of mutation is mitotic (or early meiotic prophase) DNA copying and repair error. Which may prove to be the case, but there is for example a correlation between SNPs and regions of high recombination (eg Lercher & Hurst, Trends in Genetics 18/7 pp 337-340). This would also interact with generation time - shorter generations = a higher proportion of recombinational mutation per cycle.

More dubious is the assertion that the Y study 'confirms theory'. If we have a rate of 130 mutations per (gender-not-specified) individual, I would not expect the rate on a Y, with a very individualistic approach to homologous repair and a faster turnover in the germ line, to be in the same ball park. Confirmation bias?

Anonymous said...

... and of course mutation is stochastic. As, indeed, is the actual number of cell divisions separating each sperm from its originating zygote. Further, the relationship between paternal error rate and cell divisions is not linear - older parents make even more mistakes, so 13 generations clustering around the mean would give a different mutation rate from one with the same number of replications but higher variance.

All of this makes extrapolation from the Y study, or even regarding it as confirmatory of the in vitro error rate, a little premature. 4 mutations from 13 generations of a 400-cell-cycle 10.15Mb unit is well within the bounds of expectation for mutation rates from under half to over 1.5 times the in vitro figure if the division count is right, and likewise for 2/3rds to twice the division count if the mutation rate is right, and sundry combinations thereof. And of course we are light by any mutations that were 'corrected' by selection - population size is irrelevant here; we are looking at two surviving Y chromosomes.

So yes, I think there is some work to be done before we would need to adjust divergence times!