Enhancing statistical frameworks for inferring natural selection using ancient genomes data.

Lucas Sort (RIKEN)
January 06, 2026 — 11:00 — "Salle G. Hopper 0336, IBM" (and Teams)

Abstract

Over the past decade, the emergence of ancient DNA has opened new opportunities for studying evolutionary processes. However, inferring signals of selection from such data remains a methodological challenge since controlling for confounding evolutionary processes is difficult. In this context, ancient DNA time series data, which have proliferated, have led to the development of methods based on two main frameworks: Hidden Markov Models and Generalized Linear Mixed Models. In this talk, I will introduce the statistical foundations of population genetics used in this context, and discuss how we aim to clarify the links between these frameworks for inferring selection and the classical Wright–Fisher model, enabling targeted modeling improvements and producing more relevant comparisons across methods.

Bio

Currently a postdoctoral researcher in the Mathematical Genomics Research Unit (led by Dr. Leo Speidel) at the RIKEN Center for Interdisciplinary Theoretical and Mathematical Sciences (iTHEMS) in Tokyo, Japan, I work on developing and improving statistical methods for inferring evolutionary processes such as natural selection and migrations in past human populations. I was previously a Ph.D. student at Université Paris-Saclay, from which I graduated in January 2025, after working on the development of new statistical approaches for studying longitudinal data. More generally, my work focuses on how we can apply Mathematics and Statistics to solve problems encountered in Biology and Medical Sciences.