A Genotype-phenotype Taxonomy of Hypertrophic Cardiomyopathy

Lara Curran, Antonio de Marvao, Paolo Inglese, Kathryn A. McGurk, Adam Clement, Sean L. Zheng, Surui Li, Chee Jian Pua, Mit Shah, Mina Jafari, Pantazis Theotokis, Rachel Buchan, Sean J. Jurgens, Claire E. Raphael, John Baksi, Antonios Pantazis, Brian P. Halliday, Dudley J. Pennell, Wenjia Bai, Calvin Chin, Rafik Tadros, Connie R. Bezzina, Hugh Watkins, Stuart A. Cook, Sanjay Prasad, James S. Ware, Declan P. O’Regan

This study, recently published in medRxiv, represents a significant step forward in how we classify and understand Hypertrophic Cardiomyopathy (HCM)—a leading cause of sudden cardiac death. By moving away from subjective visual assessments and toward a data-driven, 3D machine-learning approach, the researchers have mapped the complex relationship between a patient's genetic makeup and the physical structure of their heart.

TL;DR

Researchers developed a "3D taxonomy" of heart shapes in HCM patients using machine learning on cardiac MRI data. This map reveals that heart remodeling exists on a continuum where specific structural patterns are strongly linked to high-risk genetic variants and significantly higher mortality rates.

What Problem Does This Paper Address?

Hypertrophic cardiomyopathy (HCM) is notoriously "heterogeneous," meaning it looks and behaves differently in every patient. Traditionally, doctors have classified HCM based on subjective descriptions of heart wall thickening (e.g., "sigmoid" or "reverse curve").

However, these labels often fail to capture the underlying biology. Some patients have clear "sarcomeric" genetic mutations (the classic cause), while others do not, despite having similar-looking hearts. There is currently no systematic framework that links the 3D shape of the heart (phenotype) with genetic risk (genotype) and clinical outcomes. This gap makes it difficult to personalize treatment or accurately predict which patients are at the highest risk for heart failure or sudden death.

Methods

The authors employed a sophisticated multi-step pipeline:

  1. 3D Cardiac Modeling: Using deep learning, they automatically segmented Cardiac Magnetic Resonance (CMR) images to create high-resolution 3D models of the left ventricle (LV) for over 17,000 individuals.
  2. Unsupervised Learning: They used a technique called DDRTree (discriminative dimensionality reduction via learning a tree) to project these complex 3D shapes into a 2D "phenotypic tree." This allowed them to see how different heart shapes transition into one another.
  3. Genomic Integration: They stratified patients by rare variants (pathogenic/likely pathogenic, VUS, or negative) and calculated Polygenic Risk Scores (PGS) to account for common genetic modifiers.
  4. Survival Analysis: They projected mortality data onto the tree to see if specific "branches" were deadlier than others.

Datasets

  • Discovery Cohort: 436 HCM patients from the NIHR Royal Brompton Hospital Cardiovascular Biobank (UK).
  • Validation Cohort: 60 HCM patients from Singapore to test the model's generalizability across different ancestries.
  • Reference Population: 16,691 genotype-negative individuals from the UK Biobank to establish "normal" heart geometry.

Key Insights and Ideas

  • The Genotype-Phenotype Paradox: Surprisingly, patients with the most dangerous genetic variants (SARC-P/LP) actually had lower overall heart mass than those without mutations. However, their thickening was much more localized to the basal septum (the top part of the wall dividing the heart chambers).
  • Four Main "Branches" of Disease: The unsupervised model identified four distinct structural trajectories:
  • Non-sarcomeric: Thickening at the mid-to-apex, often linked to hypertension.
  • Obstructive: Diffuse thickening associated with blood flow blockage (LVOTO).
  • Isolated Basal: A milder, potentially age-related form.
  • Familial Sarcomeric: Milder overall thickening but highly enriched for high-risk genetic mutations.
  • Continuous vs. Categorical: The study argues that HCM isn't a set of discrete boxes but a continuum. The "tree" captures the transition from undifferentiated heart shapes in the center to highly specific disease patterns at the tips of the branches.

Key Results

  • Mortality Risk: Carriers of pathogenic variants (SARC-P/LP) had a 2.66x higher risk of death compared to genotype-negative patients, with a significantly shorter median lifespan (67 vs. 76 years).
  • Genetic Enrichment: Branch 4 was the most "genetic," with an odds ratio of 2.18 for carrying a pathogenic variant.
  • Generalizability: The model proved robust when applied to the Singaporean cohort, achieving a "trustworthiness" score of 0.86–0.88, suggesting the 3D patterns of HCM transcend ethnic boundaries.
  • Polygenic Influence: Even in patients without a single "smoking gun" mutation, common genetic variations (measured via PGS) significantly influenced the specific shape and severity of heart remodeling.

Limitations and Future Work

The authors acknowledge that their primary biobank may be biased toward patients stable enough to undergo MRI, potentially missing the most acute or severe cases. Additionally, while the Singaporean cohort provided some ancestral diversity, the vast majority of the training data was from individuals of European descent.

Future work aims to integrate "multi-omic" data—like proteomics and transcriptomics—directly into the 3D map. The ultimate goal is a clinical tool where a patient’s MRI can be "mapped" onto the tree to immediately provide a personalized risk profile and genetic probability.

Summary updated

Abstract

Abstract Background Hypertrophic cardiomyopathy (HCM) is an important cause of sudden cardiac death associated with heterogeneous phenotypes but there is no systematic framework for classifying morphology or assessing associated risks. Here we quantitatively survey genotype-phenotype associations in HCM to derive a data-driven taxonomy of disease expression. Methods We enrolled 436 HCM patients (median age 60 years; 28.8% women) with clinical, genetic and imaging data. An independent cohort of 60 HCM patients from Singapore (median age 59 years; 11% women) and a reference population from UK Biobank (n = 16,691, mean age 55 years; 52.5% women) were also recruited. We used machine learning to analyse the three-dimensional structure of the left ventricle from cardiac magnetic resonance imaging and build a tree-based classification of HCM phenotypes. Genotype and mortality risk distributions were projected on the tree. Results Carriers of pathogenic or likely pathogenic variants for HCM (P/LP) variants had lower left ventricular mass, but greater basal septal hypertrophy, with reduced lifespan (mean follow-up 9.9 years) compared to genotype negative individuals (hazard ratio: 2.66; 95% confidence interval [CI]: 1.42-4.96; P < 0.002). Four main phenotypic branches were identified using unsupervised learning of three-dimensional shape: 1) non-sarcomeric hypertrophy with co-existing hypertension; 2) diffuse and basal asymmetric hypertrophy associated with outflow tract obstruction; 3) isolated basal hypertrophy; 4) milder non-obstructive hypertrophy enriched for familial sarcomeric HCM (odds ratio for P/LP variants: 2.18 [95% CI: 1.93-2.28, P = 0.0001]). Polygenic risk for HCM was also associated with different patterns and degrees of disease expression. The model was generalisable to an independent cohort (trustworthiness M 1 : 0.86-0.88). Conclusions We report a data-driven taxonomy of HCM for identifying groups of patients with similar morphology while preserving a continuum of disease severity, genetic risk and outcomes. This approach will be of value in understanding the causes and consequences of disease diversity.
BibTeX
@article{curran2024genotype,
  author = {Lara Curran and Antonio de Marvao and Paolo Inglese and Kathryn A. McGurk and Adam Clement and Sean L. Zheng and Surui Li and Chee Jian Pua and Mit Shah and Mina Jafari and Pantazis Theotokis and Rachel Buchan and Sean J. Jurgens and Claire E. Raphael and John Baksi and Antonios Pantazis and Brian P. Halliday and Dudley J. Pennell and Wenjia Bai and Calvin Chin and Rafik Tadros and Connie R. Bezzina and Hugh Watkins and Stuart A. Cook and Sanjay Prasad and James S. Ware and Declan P. O’Regan},
  title = {{A Genotype-phenotype Taxonomy of Hypertrophic Cardiomyopathy}},
  journal = {Journal of Cardiovascular Magnetic Resonance},
  year = {2024},
  doi = {10.1016/j.jocmr.2024.100168},
}

Loading related content...

Figure