Transcending the Annotation Bottleneck: AI-Powered Discovery in Biology and Medicine

TL;DR

This paper reviews how unsupervised and self-supervised learning methods are transforming biomedicine by enabling discovery from large datasets without reliance on expert annotations, leading to novel insights and improved diagnostics.

Contribution

It synthesizes recent advances in unsupervised learning in biomedicine, demonstrating their ability to uncover phenotypes, link morphology to genetics, and detect pathologies without labels.

Findings

Unsupervised methods can derive heritable cardiac traits.

They can predict spatial gene expression in histology.

They can detect pathologies with performance comparable to supervised methods.

Abstract

The dependence on expert annotation has long constituted the primary rate-limiting step in the application of artificial intelligence to biomedicine. While supervised learning drove the initial wave of clinical algorithms, a paradigm shift towards unsupervised and self-supervised learning (SSL) is currently unlocking the latent potential of biobank-scale datasets. By learning directly from the intrinsic structure of data - whether pixels in a magnetic resonance image (MRI), voxels in a volumetric scan, or tokens in a genomic sequence - these methods facilitate the discovery of novel phenotypes, the linkage of morphology to genetics, and the detection of anomalies without human bias. This article synthesises seminal and recent advances in "learning without labels," highlighting how unsupervised frameworks can derive heritable cardiac traits, predict spatial gene expression in histology, and detect pathologies with performance that rivals or exceeds supervised counterparts.