The Selection Landscape and Genetic Legacy of Ancient Eurasians. Evan K Irving-Pease, Alba Refoyo-Martínez, Andrés Ingason, Alice Pearson, Anders Fischer, William Barrie, ..., Leo Speidel, ..., Peter H Sudmant, Daniel J Lawson, Richard Durbin, Thorfinn Korneliussen, Thomas Werge, Morten E Allentoft, Martin Sikora, Rasmus Nielsen, Fernando Racimo, Eske Willerslev.
Preprint: bioRxiv:2022.09.22.509027
Integrative analysis of GWAS and co-localisation data suggests novel genes associated with age-related multimorbidity. Clare E West, Mohd Karim, Maria J Falaguera, Leo Speidel, ..., Brian Marsden.
Preprint: medRxiv:2022.11.11.22282236
Balancing selection on genomic deletion polymorphisms in humans. Alber Aqil, Leo Speidel, Pavlos Pavlidis, Omer Gokcumen. ELife 12, e79111 (2023)
This paper looks into old but polymorphic deletions in the human genome, such as those that are shared with Neanderthals, and quantifies the evidence for balancing selection acting on these.
Grey wolf genomic history reveals a dual ancestry of dogs. Anders Bergström, ..., Leo Speidel, ..., Pontus Skoglund. Nature volume 607, 313–320 (2022)
This paper analyses 72 ancient wolf genomes from the past 100,000 years and among others finds that most modern-day dogs derive their ancestry from two distinct wolf sources. It also quantifies selection directly using an ancient DNA time series. We confirmed that the strongest selection hit also has the youngest time to the most recent common ancestor in the whole genome, i.e. all wolves coalesce very quickly - we can infer this using only modern-day wolves complementing the time series approach.
Multi-ancestry genetic study of type 2 diabetes highlights the power of diverse populations for discovery and translation. Anubha Mahajan, Cassandra N Spracklen*, Weihua Zhang*, Maggie CY Ng*, Lauren E Petty*, Hidetoshi Kitajima*, Grace Z Yu*, Sina Rueger*, Leo Speidel*, ..., Mark I McCarthy, Andrew P Morris. (*joint second authors) Nature Genetics 54, 560–572 (2022)
Preprint: medRxiv:2020.09.22.20198937
This paper did a GWAS meta analysis on ~1.2m subjects from diverse ancestries, identifying 338 distinct association signals for T2D. We looked at polygenic selection using Relate trees, and found some evidence of selection for increased T2D risk in 1000G African ancestry groups, which appears to be driven by a subset of SNPs that are also associated with weight/fat distribution in UKB.
Sex-specific phenotypic effects and evolutionary history of an ancient polymorphic deletion of the human growth hormone receptor. Marie Saitou, ..., Leo Speidel, ..., Omer Gokcumen. Science Advances 7, eabi4476 (2021)
This paper looks into the function and evolutionary history of the deletion of exon 3 in the growth hormone receptor. This deletion appears to be associated with protection from malnutrition and exhibits quite remarkable interactions with diet and sex in mice. This deletion is fixed in all archaics we looked at (incl. of lower coverage). Despite its apparent old age, it is segregating everywhere today and is found at especially low frequency in East Asia.
This paper extends Relate to work with ancient DNA and introduces Colate, a method for inferring coalescence rates between low-coverage (ancient) genomes. We also study geographic and temporal patterns of the TCC/TTC mutation rate change in moderns and 161 ancients, finding that it was already widespread >10k years ago and correlates with recent ancestry from a 10k year old Anatolian.
This paper introduces a full-likelihood method to quantify polygenic adaptation. The method uses Relate-sampled genealocical trees and an importance sampler similar to the CLUES method, and can be used to disentangle varying amount of selection acting on correlated traits.
We studied percolation of disks which are dropped at random onto a plane leading to clusters of overlapping disks, the size of which can undergo sudden phase transitions. Here, we characterize topological properties of such clusters.
For a simple protocol for disseminating information in a network, we derive tight bounds on the time until all nodes are informed, which we show to be robust to the density of connections in the network.
The structure of networks describing human interactions directly impact how easily epidemics can spread. Many networks additionally change through time and we show that rapidly changing networks are always more susceptible to an epidemic compared to the corresponding static ones.
Empirical networks commonly exhibit communities, which are densely connected submodules. Detecting such communities can substantially enhance the understanding of a network. We extend an existing approach to a subclass of networks that have directed links and no cycles, including e.g., citation networks and genealogies.
We characterise a random walk on a temporally changing network. This random walk is a simplified model for e.g., an epidemic spreading through physical contacts among agents.