Masked Autoencoders are Scalable Learners of Cellular Morphology

TL;DR

This paper demonstrates that large-scale self-supervised masked autoencoders, especially ViT-based models, significantly improve the inference of biological relationships from cellular microscopy images compared to previous methods.

Contribution

It shows that scaling up masked autoencoders on large microscopy datasets enhances biological signal capture, outperforming weakly supervised baselines.

Findings

ViT-L/8 trained on 3.5 billion crops outperforms baselines by up to 28%.

Self-supervised models better capture biological relationships than hand-crafted features.

Scaling models improves inference accuracy on cellular morphology data.

Abstract

Inferring biological relationships from cellular phenotypes in high-content microscopy screens provides significant opportunity and challenge in biological research. Prior results have shown that deep vision models can capture biological signal better than hand-crafted features. This work explores how self-supervised deep learning approaches scale when training larger models on larger microscopy datasets. Our results show that both CNN- and ViT-based masked autoencoders significantly outperform weakly supervised baselines. At the high-end of our scale, a ViT-L/8 trained on over 3.5-billion unique crops sampled from 93-million microscopy images achieves relative improvements as high as 28% over our best weakly supervised baseline at inferring known biological relationships curated from public databases. Relevant code and select models released with this work can be found at: https://github.com/recursionpharma/maes_microscopy.