A theoretical advantage that didn't survive contact with a real benchmark.

A theoretical advantage that didn't survive contact with a real benchmark.

Biobank-scale imaging provides an unprecedented opportunity to characterise how thousands of organ phenotypes vary in populations. However, deriving specific phenotypes from imaging data requires time-consuming expert annotation, limiting scalability. In this study, we develop a 3D diffusion autoencoder to derive latent phenotypes from temporally resolved cardiac MRI data of 71,017 UK Biobank participants. These phenotypes are reproducible, heritable (h2 = [4—18%]), and significantly associated with cardiometabolic traits. To establish the genetic basis of such traits, we perform a genome-wide association study, identifying 89 significant common variants (P < 2.3 × 10−9) across 42 loci, including seven novel loci. Extensive multi-trait colocalisation analyses (PP.H4 > 0.8) link variants across phenotypic scales, from intermediate cardiac traits to cardiac disease endpoints. In conclusion, this study showcases the use of diffusion autoencoding methods as p

Sparse autoencoders (SAEs) have emerged as a powerful tool for interpreting neural networks by extracting the concepts represented in their activations. However, choosing the size of the SAE dictionar...

Linear autoencoders (LAEs) have gained increasing popularity in recommender systems due to their simplicity and strong empirical performance. Most LAE models, including the Emphasized Denoising Linear...

The authors declare no competing interests. Ethics statement This study exclusively utilized publicly available, de-identified datasets. The METABRIC dataset was accessed via the pycox Python library, which provides publicly available de-identified data originally from the METABRIC consortium. The MIMIC-IV database was accessed through PhysioNet following completion of the required human subjects research training (CITI Program) and execution of a data use agreement (credentialed user access). Both datasets contain only de-identified patient information in accordance with the Health Insurance Portability and Accountability Act (HIPAA) Safe Harbor provisions. No new data collection, patient contact, or prospective intervention was involved in this study. As this research involved only secondary analysis of existing de-identified data, no additional ethical approval was required beyond the data use agreements mandated by the data repositories. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence,...

The findings position phase-aware functional autoencoders as a corrective to linear reduction methods that conflate timing shifts with true shape differences.