Interpretable Phenotype Discovery Using Temporal Risk-Aware Clustering
- 1Georgia Institute of Technology
Abstract
Longitudinal clinical cohorts contain heterogeneous progression patterns that are obscured by coarse diagnostic labels. Discovering these latent progression phenotypes requires models that distinguish patients not only by their outcomes, but also by how their clinical trajectories evolve. This paper proposes TRACE, an interpretable temporal phenotyping method that enables the learning of outcome-informed progression phenotypes from longitudinal trajectories. TRACE combines a trajectory-and outcome-aware contrastive objective with a probabilistic predictive phenotype layer, encouraging phenotypes to capture progression structure while remaining predictive of diagnostic state. A dual-attention mechanism further supports interpretation by identifying the visits and clinical variables most associated with each phenotype assignment. TRACE is evaluated on the Alzheimers Disease Neuroimaging Initiative (ADNI) cohort for cognitive disease modeling, showing improved performance in both diagnostic prediction and phenotype discovery compared with other temporal phenotyping baselines. Finally, by allowing the number of phenotypes to exceed the diagnostic labels, TRACE enables discovering additional progression patterns that are not captured by labels defined a priori.
Related
- Temporal Clustering of Acute Neurological Disorders: Testing the Clinical Impression of Diagnostic 'Theme Shifts'
- KT-CBHNET: a spatio-temporal deep learning framework for interpretable ECG signal classification for cardiovascular disease prediction
- Expert knowledge-based weighted K-means clustering for polycystic ovary syndrome phenotype identification
- A leakage-aware and interpretable machine learning framework for ovarian cancer prediction using the PLCO screening dataset
- Trial After Trial: Temporal Dynamics Reveal a Distinct Computational Phenotype of Chronic Cannabis Use
BibTeX
@misc{roudbari2026interpretable,
title = {{Interpretable Phenotype Discovery Using Temporal Risk-Aware Clustering}},
author = {A. Roudbari and P. Van Hentenryck and R. Zandehshahvar},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.18.26365169},
url = {https://www.medrxiv.org/content/10.64898/2026.10.18.26365169v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Interpretable Phenotype Discovery Using Temporal Risk-Aware Clustering
AU - A. Roudbari
AU - P. Van Hentenryck
AU - R. Zandehshahvar
PY - 2026
DA - 2026/10/18
PB - medRxiv
DO - 10.64898/2026.10.18.26365169
UR - https://www.medrxiv.org/content/10.64898/2026.10.18.26365169v1
AB - Longitudinal clinical cohorts contain heterogeneous progression patterns that are obscured by coarse diagnostic labels. Discovering these latent progression phenotypes requires models that distinguish patients not only by their outcomes, but also by how their clinical trajectories evolve. This paper proposes TRACE, an interpretable temporal phenotyping method that enables the learning of outcome-informed progression phenotypes from longitudinal trajectories. TRACE combines a trajectory-and outcome-aware contrastive objective with a probabilistic predictive phenotype layer, encouraging phenotypes to capture progression structure while remaining predictive of diagnostic state. A dual-attention mechanism further supports interpretation by identifying the visits and clinical variables most associated with each phenotype assignment. TRACE is evaluated on the Alzheimers Disease Neuroimaging Initiative (ADNI) cohort for cognitive disease modeling, showing improved performance in both diagnostic prediction and phenotype discovery compared with other temporal phenotyping baselines. Finally, by allowing the number of phenotypes to exceed the diagnostic labels, TRACE enables discovering additional progression patterns that are not captured by labels defined a priori.
ER -