Kieran Elmes

Accuracy and Scalability of Machine Learning Methods for Genotype-Phenotype Association Data

This study compares machine learning methods for predicting phenotypes from genetic data, finding that transformers outperform others in simulations and match state-of-the-art models on real data.

A Fast Lasso-Based Method for Inferring Higher-Order Interactions

The paper introduces a fast Lasso-based method, Pint, for inferring three-way gene interactions in large-scale genetic data, outperforming existing methods in simulated and real datasets.

SNVformer: An Attention-based Deep Neural Network for GWAS Data

This paper introduces SNVformer, a Transformer-based model for GWAS data that predicts gout from SNVs with 83% AUROC, showing promise for scalable genotype-phenotype association analysis.

Learning epistatic gene interactions from perturbation screens

The paper develops a model to infer pairwise epistatic gene interactions from siRNA perturbation screens, showing that glinternet outperforms xyz in accuracy for high-dimensional data.

A Fast Lasso-based Method for Inferring Pairwise Interactions

The paper introduces a fast Lasso-based method for detecting epistatic interactions in large-scale genotype-phenotype screens, demonstrating its application on simulated and real data.

Geometry of Ranked Nearest Neighbour Interchange Space of Phylogenetic Trees

This paper analyzes the geometry of the ranked nearest neighbour interchange graph for phylogenetic trees, finding exact radius and diameter values and comparing it to the classical NNI graph.