Efficient Reconstruction of Stochastic Pedigrees
This addresses the pedigree reconstruction problem for applications in genomics, with conceptual implications for genomic privacy, but it is incremental as a prototype for further investigation.
The paper tackles the problem of reconstructing genealogies from genetic data by introducing the Rec-Gen algorithm, which provides accurate reconstruction for a large fraction of pedigrees with relatively low sample complexity in terms of sequence length.
We introduce a new algorithm called {\sc Rec-Gen} for reconstructing the genealogy or \textit{pedigree} of an extant population purely from its genetic data. We justify our approach by giving a mathematical proof of the effectiveness of {\sc Rec-Gen} when applied to pedigrees from an idealized generative model that replicates some of the features of real-world pedigrees. Our algorithm is iterative and provides an accurate reconstruction of a large fraction of the pedigree while having relatively low \emph{sample complexity}, measured in terms of the length of the genetic sequences of the population. We propose our approach as a prototype for further investigation of the pedigree reconstruction problem toward the goal of applications to real-world examples. As such, our results have some conceptual bearing on the increasingly important issue of genomic privacy.