Steven Squires

The paper introduces a masked-MLP neural network to improve genetic risk score calculation from noisy and missing data, showing better robustness than standard methods.

This paper demonstrates that deep learning can generate polygenic risk scores with high accuracy, even with missing data, and shows that a single model can predict multiple scores.

This paper simulates genetic risk scores using summary statistics to enable comparison across datasets, showing similar performance to real data in type 1 diabetes.