📣Online NOW!
📄Estimation of demography and mutation rates from one million haploid genomes
🧑🤝🧑 @jgschraiber.bsky.social @jeffspence.github.io @docedge.bsky.social
Amazing, congratulations!!
To close, I’ll highlight is that our model gives a general prediction of germline mutation rate as a function of generation time. When trio-based estimates aren’t available, this model may perform better than old phylogenetic estimates. www.nature.com/articles/nrg...
We did utilize the same data generated by Wang, et al. but looked at a different aspect of the data to reach the same conclusion, which even holds when looking only at offspring of younger mothers and replicates in an independent aye-aye dataset.
Consistent with prior phasing results, we estimated that most species have a higher paternal than maternal mutation rate. However, aye-aye was the exception, corroborating the recent finding of extraordinary maternal bias in the offspring of aged aye-aye mothers! journals.plos.org/plosbiology/...
For example, since the human maternal mutation rate per year is about 1/3 of the paternal mutation rate, a 3-to-1 weighted average of paternal and maternal age has a higher correlation with trio mutation load than either the paternal or maternal age alone.
To regress the total mutation rate against average parental age, we realized that we should find the weighted average of maternal and paternal age that has the highest correlation with mutation rate. This weighting is an estimate of the ratio of maternal to paternal age effects!
We thought a lot about how to regress mutation rate against parental age. Paternal and maternal age effects are usually inferred using just phased mutations, but since only about 1/3 of mutations can be phased to a parent in short-read trios, this throws away a lot of data.
We think that this main theoretical result will be very useful for interpreting and predicting mutation rate variation in multicellular species. I also want to highlight a few of the paper’s “side results” that we think may be just as useful to the community.
To explain these results, we show that selection against a mutator allele increasing the mutation rate per year in the germ cells is likely strongest in species with long generation times, since the allele will have more time to create extra mutations during each generation.
Consistent with prior work, the mutation rate per generation is highest in long-lived species with small effective population sizes. However, the opposite is true of the mutation rate per year in the germ cell lineages after puberty! (aye-aye outlier discussed more later in this thread)
We meta-analyzed variation in germline mutation rates among species and reproductive ages in order to disentangle changes molecular parameters like DNA repair from variation caused by demographic parameters like spermatogonial and oocyte aging.
Species with long lifespans tend to have low effective population sizes because they reproduce slowly and use many environmental resources. This could lessen the effectiveness of weak selection to avoid a small number of germline mutations per generation. www.annualreviews.org/content/jour...
We wanted to understand whether species with longer lifespans have better or worse DNA repair than species with short lifespans. Longevity appears to be associated with lower somatic mutation rates, but some population genetic theory predicts the opposite trend in the germline 🤔
. @pnas.org just published the final version of our manuscript on how generation time and effective population size interact to shape vertebrate germline mutation rates, led by Luke Zhu and Annabel Beichman: www.pnas.org/doi/10.1073/...
Whoops, used the wrong handle to tag @aaronquinlan.bsky.social above!
We’ve released this simple, versatile tool as an R package and we hope that others will try it out as a simple aid for the interpretation of mutational signature differences among tumors. All feedback is appreciated! github.com/sfhart33/mut...
Many of these associations have been previously reported, but the AMSD provides a simple, unified framework for ranking their significance and easily assessing which tumors appear to have etiologies that differ the most dramatically among populations.
We found that PolE hypermutation is most common in endometrial and colorectal tumors from people of East Asian descent, lung cancers from African Americans have more tobacco-associated mutations, and melanomas from Europeans have a higher UV-associated mutation load.
Next, we tested whether genetic ancestry was associated with human tumor mutation spectra. The AMSD revealed a suite of associations between ancestry and mutation load, which could reflect both genetic differences in cancer susceptibility and social determinants of health.
We first used the AMSD to reassess a previous study reporting that many carcinogens failed to cause distinctive mutational signatures in a mouse experiment. We found that many of these carcinogens still significantly perturbed the composition of the mutation spectrum.
We now generalize the AMSD to test which exposures appear to perturb tumor mutation spectra. If this test finds that two groups of tumors have statistically indistinguishable mutation spectra, that saves us overintepreting small differences in their mutational signature profiles.
We previously developed the AMSD with Tom Sasani and @aaronquinlan to test whether two groups of sequences have significantly different mutation spectra. We successfully used it to identify a germline mutator allele in the BXD mice: elifesciences.org/articles/89096
Mutational signature analysis is a very powerful tool for decomposing tumor mutation burdens into different endogenous and exogenous exposures, but when underlying mutation counts are sparse, it is not easy to tell which differences between signature profiles are meaningful.
If you’ve ever wondered about the statistical significance of differences among mutational signature profiles, check out our new Aggregate Mutation Spectrum Distance (AMSD) preprint co-led by Sam Hart and @alisonfeder.bsky.social with @nalcala.bsky.social www.biorxiv.org/cgi/content/...
