Deep phenotyping of heart failure with preserved ejection fraction through multi-omics integration

Area of Science:

• Cardiology
• Genomics
• Computational Biology

Background:

• Heart failure with preserved ejection fraction (HFpEF) is the most common type of heart failure, posing significant challenges due to its heterogeneous nature and limited treatment options.
• Early identification of HFpEF is critical for implementing effective preventive strategies, especially considering its association with lifestyle-related comorbidities.

Purpose of the Study:

• To introduce and validate a machine learning-based multi-omics approach for early detection and characterization of HFpEF.
• To integrate clinical and molecular data for improved diagnostic and prognostic capabilities in HFpEF.

Main Methods:

• A supervised classifier was trained on a large UK Biobank cohort (n=401,917) using phenotypic profiles.
• Model performance was validated in an independent hold-out set (n=100,946) including individuals with and without HFpEF.
• Similarity Network Fusion (SNF) was employed to identify distinct HFpEF subgroups based on multi-omics data.

Main Results:

• The classifier achieved high accuracy in identifying HFpEF phenotypes, with an ROC AUC of 0.931.
• The model detected individuals who later developed HFpEF an average of 6.3 years prior to symptom onset.
• A high-risk HFpEF subgroup with elevated mortality and inflammatory pathway dysregulation was identified with an ROC AUC of 0.988.

Conclusions:

• The study successfully identified HFpEF phenotypes at an early, potentially modifiable stage, years before clinical manifestation.
• The multi-omics approach provides novel insights into HFpEF complexity and enables more precise risk stratification.
• Early detection and characterization are key to improving preventive strategies and patient outcomes for HFpEF.