Data science in history and health: Polynesia and polynucleotides
Consisting of millions to billions of linked features, genomic datasets from both living and ancient humans are rapidly growing to sample sizes of hundreds of thousands of individuals, and within each of these samples lie the genomic fragments of far vaster numbers of ancestors. This data offers exciting opportunities both to reconstruct the past, and, together with electronic medical records, to illuminate human health and disease. I will discuss two such datasets that we've assembled, from Latin America and from Oceania, and cover the methods we've introduced for genetic assignment (unsupervised and supervised), for dating, for dimensionality reduction with missing data, and for uncovering and reconstructing the telescoping series of founder effects that peopled Polynesia. In the process, we'll answer the question that Captain Cook, FRS, first posed two and a half centuries ago: "How shall we account for this nation spreading itself so far over this vast ocean?"