Sara Mathieson

Interpreting Convolutional Neural Networks in Population Genetics

This study investigates how convolutional neural networks in population genetics implicitly compute summary statistics and reveals their ability to efficiently approximate long-range linkage disequilibrium.

A reusable neural approach to recombination mapping for model and non-model species

The paper introduces fastrho, a neural network for recombination mapping that works across species and scenarios, showing it can produce accurate recombination maps and detect selection effects.

Popformer: Learning general signatures of positive selection with a self-supervised transformer

Popformer is a transformer-based model pre-trained on human genomic data to detect selection signatures, showing improved accuracy in classification and generalization to real data.

Fast pairwise coalescence enables gene-resolution scans for recent selection in diverse human populations

This paper introduces a GPU-accelerated method to compute pairwise coalescence times at gene resolution, identifying recent selection signals in diverse human populations.

On the use of generative models for evolutionary inference of malaria vectors from genomic data

This paper uses generative deep learning to infer the evolutionary history of malaria vectors in West Africa, finding that a migration model better explains population dynamics than a model without post-split migration.

Identity-by-descent (IBD) segment outlier detection in endogamous populations using pedigree cohorts

The paper proposes an IBD pruning algorithm using pedigree information to reduce false positives in IBD segment detection for endogamous populations.

Interpreting Generative Adversarial Networks to Infer Natural Selection from Genetic Data

This paper introduces a GAN-based method to detect natural selection by training on neutral data and identifying regions that deviate from simulated neutral patterns.

The genomic footprint of social stratification in admixing American populations

This study infers sociocultural stratification in admixed American populations using genomic data and a mating model with ancestry-related assortative mating and sex bias parameters.

Automatic inference of demographic parameters using Generative Adversarial Networks

This paper introduces pg-gan, a GAN-based method to estimate demographic parameters in population genetics, showing it can recover input parameters in simulations and recapitulate real human data features.

Ancestral Haplotype Reconstruction in Endogamous Populations using Identity-By-Descent

This paper introduces an algorithm called thread to reconstruct ancestral haplotypes in endogamous populations using pedigree and genotype data, finding it can reconstruct 79% of haplotypes for 224 ancestral individuals per chromosome.