Daniel J. Lawson’s research while affiliated with University of Bristol and other places

Genome-wide association studies (GWAS) have revolutionized our understanding of the genetic basis of complex traits and diseases, but limitations in SNP-centric approaches to population stratification limit the resolution of fine-scale population structures. Here we consider the use of haplotypes to represent population structure, leveraging haplotype components (HCs) for an improved understanding of trait associations and adjustment for population stratification. Using data from the UK Biobank, we showed that HCs have stronger associations with a range of phenotypes than principal components (PCs) while containing more predictive power for birthplaces globally. In GWAS, HCs-correction identifies more genome-wide significant association signals for birthplace- and lifestyle-related phenotypes, which are missed by PCs-corrected GWAS. Through thorough testing and simulation, we highlight challenges in performing ancestry-specific GWAS, underscoring the critical role of accurate local ancestry inference in studying admixed populations. We analyzed the haplotype structure of the UK Biobank in terms of 93 genetically-distinct populations, which enabled the computation of Ancestral Risk Scores (ARS) across 8 continental populations, providing insights into population-specific genetic risks for traits and diseases. By integrating haplotype information, this framework provides the potential to address challenges in population stratification, enhances GWAS resolution, and supports equitable health research by facilitating genetic studies in diverse populations.

This paper employs Unfolded Adjacency Spectral Embedding (UASE) to investigate the temporal evolution of economic relationships between locations in Great Britain. We utilise timestamped, geolocated website hyperlinks data between archived, commercial websites in Britain, which are aggregated to create a set of directed, weighted networks of hyperlink connections between Local Authority Districts (LADs) for each year in the period 2005-2010. Thus, we are able to assess the digital evolution of longstanding economic disparities such as the North-South, Urban-Rural, and London versus the rest of the country divides in Britain. Our method is a robust and scalable statistical testing procedure for detecting changes between communities in dynamic networks where changes are expressed in terms of known covariates. We can describe network trends over time with respect to longitude and latitude covariates, relying on spatio-temporal stability properties of UASE to make comparisons of nodes in graphs over time. These trends can be formally tested with p -values as well as interpreted in terms of covariates and features of the network. We show how the methodology can be made robust to the problems of large-scale real-world data, used to detect changes over time, and identify their characteristics. This work provides the first robust evidence that commercial website hyperlink connectivity patterns between the North and South are diverging over time, highlighting an increasing digital divide.

Stability for dynamic network embeddings ensures that nodes behaving the same at different times receive the same embedding, allowing comparison of nodes in the network across time. We present attributed unfolded adjacency spectral embedding (AUASE), a stable unsupervised representation learning framework for dynamic networks in which nodes are attributed with time-varying covariate information. To establish stability, we prove uniform convergence to an associated latent position model. We quantify the benefits of our dynamic embedding by comparing with state-of-the-art network representation learning methods on three real attributed networks. To the best of our knowledge, AUASE is the only attributed dynamic embedding that satisfies stability guarantees without the need for ground truth labels, which we demonstrate provides significant improvements for link prediction and node classification.

Understanding genetic differences between populations is essential for avoiding confounding in genome-wide association studies and improving polygenic score (PGS) portability. We developed a statistical pipeline to infer fine-scale Ancestry Components and applied it to UK Biobank data. Ancestry Components identify population structure not captured by widely used principal components, improving stratification correction for geographically correlated traits. To estimate the similarity of genetic effect sizes between groups, we developed ANCHOR, which estimates changes in the predictive power of an existing PGS in distinct local ancestry segments. ANCHOR infers highly similar (estimated correlation 0.98 ± 0.07) effect sizes between UK Biobank participants of African and European ancestry for 47 of 53 quantitative phenotypes, suggesting that gene–environment and gene–gene interactions do not play major roles in poor cross-ancestry PGS transferability for these traits in the United Kingdom, and providing optimism that shared causal mutations operate similarly in different populations.

Roman writers found the relative empowerment of Celtic women remarkable¹. In southern Britain, the Late Iron Age Durotriges tribe often buried women with substantial grave goods². Here we analyse 57 ancient genomes from Durotrigian burial sites and find an extended kin group centred around a single maternal lineage, with unrelated (presumably inward migrating) burials being predominantly male. Such a matrilocal pattern is undescribed in European prehistory, but when we compare mitochondrial haplotype variation among European archaeological sites spanning six millennia, British Iron Age cemeteries stand out as having marked reductions in diversity driven by the presence of dominant matrilines. Patterns of haplotype sharing reveal that British Iron Age populations form fine-grained geographical clusters with southern links extending across the channel to the continent. Indeed, whereas most of Britain shows majority genomic continuity from the Early Bronze Age to the Iron Age, this is markedly reduced in a southern coastal core region with persistent cross-channel cultural exchange³. This southern core has evidence of population influx in the Middle Bronze Age but also during the Iron Age. This is asynchronous with the rest of the island and points towards a staged, geographically granular absorption of continental influence, possibly including the acquisition of Celtic languages.

Quantifying uncertainty in networks is an important step in modelling relationships and interactions between entities. We consider the challenge of bootstrapping an inhomogeneous random graph when only a single observation of the network is made and the underlying data generating function is unknown. We utilise an exchangeable network test that can empirically validate bootstrap samples generated by any method, by testing if the observed and bootstrapped networks are statistically distinguishable. We find that existing methods fail this test. To address this, we propose a principled, novel, distribution-free network bootstrap using k-nearest neighbour smoothing, that can regularly pass this exchangeable network test in both synthetic and real-data scenarios. We demonstrate the utility of this work in combination with the popular data visualisation method t-SNE, where uncertainty estimates from bootstrapping are used to explain whether visible structures represent real statistically sound structures.

Classical point process models, such as the epidemic-type aftershock sequence (ETAS) model, have been widely used for forecasting the event times and locations of earthquakes for decades. Recent advances have led to Neural Point Processes (NPPs), which promise greater flexibility and improvements over classical models. However, the currently-used benchmark dataset for NPPs does not represent an up-to-date challenge in the seismological community since it lacks a key earthquake sequence from the region and improperly splits training and testing data. Furthermore, initial earthquake forecast benchmarking lacks a comparison to state-of-the-art earthquake forecasting models typically used by the seismological community. To address these gaps, we introduce EarthquakeNPP: a collection of benchmark datasets to facilitate testing of NPPs on earthquake data, accompanied by a credible implementation of the ETAS model. The datasets cover a range of small to large target regions within California, dating from 1971 to 2021, and include different methodologies for dataset generation. In a benchmarking experiment, we compare three spatio-temporal NPPs against ETAS and find that none outperform ETAS in either spatial or temporal log-likelihood. These results indicate that current NPP implementations are not yet suitable for practical earthquake forecasting. However, EarthquakeNPP will serve as a platform for collaboration between the seismology and machine learning communities with the goal of improving earthquake predictability.

The rapid growth of earthquake catalogs, driven by machine learning-based phase picking and denser seismic networks, calls for the application of a broader range of models to determine whether the new data enhances earthquake forecasting capabilities. Additionally, this growth demands that existing forecasting models efficiently scale to handle the increased data volume. Approximate inference methods such as inlabru, which is based on the Integrated nested Laplace approximation, offer improved computational efficiencies and the ability to perform inference on more complex point-process models compared to traditional MCMC approaches. We present SB-ETAS: a simulation based inference procedure for the epidemic-type aftershock sequence (ETAS) model. This approximate Bayesian method uses sequential neural posterior estimation (SNPE) to learn posterior distributions from simulations, rather than typical MCMC sampling using the likelihood. On synthetic earthquake catalogs, SB-ETAS provides better coverage of ETAS posterior distributions compared with inlabru. Furthermore, we demonstrate that using a simulation based procedure for inference improves the scalability from O(n2)$\mathcal {O}(n^2)$ to O(nlogn)$\mathcal {O}(n\log n)$. This makes it feasible to fit to very large earthquake catalogs, such as one for Southern California dating back to 1981. SB-ETAS can find Bayesian estimates of ETAS parameters for this catalog in less than 10 h on a standard laptop, a task that would have taken over 2 weeks using MCMC. Beyond the standard ETAS model, this simulation based framework allows earthquake modellers to define and infer parameters for much more complex models by removing the need to define a likelihood function.

Graph neural networks (GNNs) are powerful black-box models which have shown impressive empirical performance. However, without any form of uncertainty quantification, it can be difficult to trust such models in high-risk scenarios. Conformal prediction aims to address this problem, however, an assumption of exchangeability is required for its validity which has limited its applicability to static graphs and transductive regimes. We propose to use unfolding, which allows any existing static GNN to output a dynamic graph embedding with exchangeability properties. Using this, we extend the validity of conformal prediction to dynamic GNNs in both transductive and semi-inductive regimes. We provide a theoretical guarantee of valid conformal prediction in these cases and demonstrate the empirical validity, as well as the performance gains, of unfolded GNNs against standard GNN architectures on both simulated and real datasets.