Mutation is the ultimate source of genetic variation. The most direct and unbiased method of studying spontaneous mutations is via mutation accumulation (MA) lines. Until recently, MA experiments were limited by the cost of sequencing and thus provided us with small numbers of mutational events and therefore imprecise estimates of rates and patterns of mutation. We used whole-genome sequencing to identify nearly 1,000 spontaneous mutation events accumulated over ∼311,000 generations in 145 diploid MA lines of the budding yeast Saccharomyces cerevisiae. MA experiments are usually assumed to have negligible levels of selection, but even mild selection will remove strongly deleterious events. We take advantage of such patterns of selection and show that mutation classes such as indels and aneuploidies (especially monosomies) are proportionately much more likely to contribute mutations of large effect. We also provide conservative estimates of indel, aneuploidy, environment-dependent dominant lethal, and recessive lethal mutation rates. To our knowledge, for the first time in yeast MA data, we identified a sufficiently large number of single-nucleotide mutations to measure context-dependent mutation rates and were able to (i) confirm strong AT bias of mutation in yeast driven by high rate of mutations from C/G to T/A and (ii) detect a higher rate of mutation at C/G nucleotides in two specific contexts consistent with cytosine methylation in S. cerevisiae.neighbor-dependent mutation rate | strongly deleterious mutation S pontaneous mutations are the source of all genetic variation in nature. The rate of emergence of new mutations and the relative proportions of advantageous, neutral, and deleterious mutations are key determinants in how species evolve and adapt to new selective challenges. Unfortunately, our knowledge of the properties of spontaneous mutations remains incomplete primarily due to the difficulty of observing large enough numbers of mutational events in an unbiased way.Analyzing patterns of divergence in nonfunctional sequences is a statistically powerful method used to study relative rates of different mutation classes. This method is applicable to most organisms and now can generally be carried out on a genomewide scale. However, this approach relies crucially on the assumption that mutations in certain regions, such as pseudogenes or fourfold degenerate codon positions, are not affected by selection and are thus reliable approximations of true mutation rate. It is now becoming apparent that selection or selectionlike processes, such as biased gene conversion, are acting even at these sequences and can substantially bias the observed patterns (1-5).Studies focusing on mutations in reporter genes use a more restrictive method that can be applied only in model organisms. In some cases, such reporter genes can be placed genome-wide and thus provide estimates of genomic variation in mutation rates. However, this approach is limited by the inability to detect mutations without a visible phenotype and thus ...