William DeWitt @wsdewitt.github.io · May 20 We next analyzed the new high-coverage 1000 Genomes data from NY Genome Center (26 human populations). We inferred a triplet mutation spectrum history for each population, and used non-negative tensor factorization to define a small vocabulary of mutation signatures. 5 =👍 (10/n)
1 likes 1 replies
?
Replies
From this decomposition we can do some projections to see that mutation histories cluster by population and by mutation type, with some spectrum components as outliers, indicating they’re doing something different over time, or over populations, or both. (11/n)