Most SNPs in the human genome are biallelic; however, there are some sites that are triallelic. We show here that there are approximately twice as many triallelic sites as we would expect by chance. This excess does not appear to be caused by natural selection or mutational hotspots. Instead we propose that a new mutation can induce another mutation either within the same individual or subsequently during recombination. We provide evidence for this model by showing that the rarer two alleles at triallelic sites tend to cluster on phylogenetic trees of human haplotypes. However, we find no association between the density of triallelic sites and the rate of recombination, which leads us to suggest that triallelic sites might be generated by the simultaneous production of two new mutations within the same individual on the same genetic background. Under this model we estimate that simultaneous mutation contributes similar to 3% of all distinct SNPs. We also show that there is a twofold excess of adjacent SNPs. Approximately half of these seem to be generated simultaneously since they have identical minor allele frequencies. We estimate that the mutation of adjacent nucleotides accounts for a little less than 1% of all SNPs.

The mutation rate is known to vary between adjacent sites within the human genome as a consequence of context, the most well-studied example being the influence of CpG dinucelotides. We investigated whether there is additional variation by testing whether there is an excess of sites at which both humans and chimpanzees have a single-nucleotide polymorphism ( SNP). We found a highly significant excess of such sites, and we demonstrated that this excess is not due to neighbouring nucleotide effects, ancestral polymorphism, or natural selection. We therefore infer that there is cryptic variation in the mutation rate. However, although this variation in the mutation rate is not associated with the adjacent nucleotides, we show that there are highly nonrandom patterns of nucleotides that extend similar to 80 base pairs on either side of sites with coincident SNPs, suggesting that there are extensive and complex context effects. Finally, we estimate the level of variation needed to produce the excess of coincident SNPs and show that there is a similar, or higher, level of variation in the mutation rate associated with this cryptic process than there is associated with adjacent nucleotides, including the CpG effect. We conclude that there is substantial variation in the mutation that has, until now, been hidden from view.

We have previously shown that there is an excess of sites that are polymorphic at orthologous positions in humans and chimpanzees and that this is most likely due to cryptic variation in the mutation rate. We showed that this might be a consequence of complex context effects since we found significant heterogeneity in triplet frequencies around coincident single nucleotide polymorphism ( SNP) sites. Here, we show that the heterogeneity in triplet frequencies is not specifically associated with coincident SNPs but is instead driven by base composition bias around CpG dinucleotides. As a result, we suggest that cryptic variation in the mutation rate is truly cryptic, in the sense that the mutation rate does not appear to depend on any specific primary sequence context. Furthermore, we propose that the patterns around CpG dinucleotides are driven by the mutability of CpG dinucleotides in different DNA contexts. We also show that the genomic distribution of coincident SNPs is nonuniform and that there are some subtle differences between the distributions of single and coincident SNPs. Furthermore, we identify regions that contain high numbers of coincident SNPs and suggest that one in particular, a region containing the gene PRIM2, may be under balancing selection.

The processes that underlie point mutations in the human genome are largely unknown.
However, the cumulative effect of these processes have a large impact on how mutation
rates vary across a number of different scales and contexts, and consequently guide our
understanding of human disease and evolution. Although variation in the mutation rate
has been characterized on many different levels, it is not fully understood the extent to
which the rate of mutation can vary outside of the general patterns already observed.
Beginning with the human genome project, many studies have produced large unbiased
sequence datasets within a number of human populations. To that end, we analysed a
number of sequence datasets in an attempt to better understand the patterns and causes
of variation in the rate of mutation that exists across the genome. Firstly, we find that
the mutation rates of single sites vary by more than is currently understood, and that this
variation is not associated with any specific process or feature on either a local or
genomic scale. Although we have been unable to uncover the source of such variation,
understanding the range of mutability at sites in the human genome is important since it
may point to functional regions, disease phenotypes and prompt further ideas on the
underlying mechanisms associated with such a result. Furthermore, we find evidence
that a mutational process that can generate the simultaneous production of two new
alleles within the same individual during a single, or tightly linked series of mutation
events increases the number of tri-allelic sites in the human genome. There are a
number of potential mechanisms that may drive this process, and the consequences of
such an event may be far reaching, as the generation of two new alleles at a single site
in functional regions may allow a more rapid exploration of evolutionary space.
Furthermore, this process appears to make a reasonable contribution to variation in the
human genome, thus providing a substrate for evolutionary change. Finally, we observe
significant variation in the mutation rate over all scales in cancer genomes. Part of this result can be explained by the actions of specific carcinogens, however it is striking that
patterns of mutation can be both consistent across different cancer types, but also very
different between individuals with the same type of cancer over different scales. This
result points to the idea that the patterns of mutation may vary widely between different
genomes under different conditions, and the identification of general patterns in a small
number of samples may not fully describe the extent to which mutation rates can vary.
Taken together, these conclusions suggest that the patterns and processes underlying
mutation are highly complex, and require further analysis if they are to be fully
understood.