Sum-h2, enabling genetic discovery for deep learning-derived phenotypes through a fast evaluation framework and arena of performance
In the recent growing interest of AI research toward biology, genetic association studies of AI- derived phenotypes from high-content modalities such as images emerges as a powerful means for biological discovery. However, such AI-phenotyping methods still lacks a good optimization target and an efficient evaluation framework. The number of discovered loci was the major criterion for evaluating…
A new framework called sum-h2 has been introduced to accelerate genetic discovery in AI-driven phenotyping from high-content modalities, such as images. AI-phenotyping methods often lack a suitable optimization target and an efficient evaluation framework. Traditional genome-wide association studies (GWAS) and subsequent loci-clumping are time-consuming and computationally intensive, hindering rapid progress in deep learning algorithms.
Sum-h2 is a 1000x faster and lightweight alternative to conventional GWAS frameworks for assessing genetic discovery via total heritability. The proposed measure, the sum of heritability over phenotypic Principal Components (PCs), is equivalent to the previously suggested linear transformation-invariant tr(P^(-1) G) through both theoretical proof and simulations.
This equivalence enables rapid estimation of sum-h2 with minimal information loss, even in samples with increased relatedness, while maintaining the relative ranking of endophenotypes based on GWAS locus counts.
To demonstrate the utility of sum-h2, researchers utilized it to establish a Genetic Discovery Arena. This arena facilitates rapid and fair comparisons among various deep learning-derived phenotyping methods, fostering advancements in the field.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
