Peter Pfaffelhuber

The paper evaluates methods to classify individuals into populations using ancestry informative markers, finding a naive Bayes classifier performs best in identifying out-of-reference individuals.

The paper improves intra-European ancestry prediction by selecting informative genetic markers using machine learning, achieving higher accuracy than existing marker panels.

This paper benchmarks forensic ancestry prediction tools against TabPFN, showing TabPFN outperforms existing methods in accuracy and ROC AUC.

The paper shows that maximum likelihood estimates in the admixture model are non-unique and provides an algorithm to compute the set of equally likely estimates.

This paper improves BGA inference methods to account for recent admixture, showing improved accuracy and a likelihood ratio test for detecting recent admixture in genotype data.

The paper develops a supervised feature selection method to identify Ancestry Informative Markers (AIMs) for Biogeographical Ancestry inference, showing improved performance over traditional methods.

The paper analyzes the effect of negative and positive feedback on gene expression noise, showing that negative feedback reduces the Fano factor by at most 2, while positive feedback can lead to unbounded noise.