Back

Hierarchical Clustering with an Ensemble of Principle Component Trees for Interpretable Patient Stratification

Pfeifer, B.; Bloice, M. D.; Sirocchi, C.; Loecher, M.

2025-02-12 health informatics
10.1101/2025.02.10.25321980 medRxiv
Show abstract

Patient stratification plays a crucial role in personalized medicine by identifying distinct subgroups of patients based on their molecular and/or clinical characteristics. However, many unsupervised machine learning-based stratification techniques fail to identify the essential biomarker traits associated with each patient group. In this paper, we present a novel approach for interpretable patient stratification using hierarchical ensemble clustering. Our method leverages feature sampling in conjunction with principal component analysis (PCA) to capture the most significant patterns and contributing biomarkers. We demonstrate the effectiveness of our approach using machine learning benchmark datasets and real-world data from The Cancer Genome Atlas (TCGA), showcasing the improved interpretability of the detected patient clusters.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.