Nathaniel S. Pope

pg_gpu accelerates population genetics statistics on GPUs, achieving significant speedups over CPU implementations while maintaining accuracy.

This paper introduces neural posterior estimation for population genetics, showing it provides accurate and efficient posterior distributions compared to existing methods.

The paper introduces cxt, a transformer-based model that predicts coalescence times from mutation patterns, achieving performance comparable to MCMC methods and enabling efficient inference on genomic data.

The paper develops a method to estimate variance components and predict genetic values using ancestral recombination graphs (ARGs) and linear mixed models on large datasets.

This paper introduces TRAMA, a method to estimate mutation models for tandem repeats using ancestral recombination graphs, showing accurate parameter estimation under SMM and TPM.

The paper develops a scalable method to estimate allele ages from large genomic datasets, revealing purifying selection and heterogeneity in ultra-rare variants.

The paper extends the stdpopsim simulation framework to include realistic selection models, enabling studies of selection's impact on genetic variation across species.

The paper introduces a method to insert non-coalescing haplotypes into ancestral recombination graphs (ARGs) and proposes new metrics to compare ARGs based on haplotype relationships.

The paper expands the stdpopsim species catalog and improves genome simulations by adding non-crossover recombination and species-specific annotations, making simulations more realistic for non-model organisms.

The paper introduces msprime 1.0, a high-performance simulation tool for ancestry and mutation in population genetics, showing it is faster and more memory efficient than existing alternatives.