Masked Autoencoders are Scalable Learners of Cellular Morphology
This work addresses the challenge of capturing biological signal from microscopy data for researchers in biology and drug discovery, representing an incremental advance in scaling existing methods.
The paper tackled the problem of inferring biological relationships from cellular phenotypes in microscopy screens by scaling self-supervised masked autoencoders on large datasets, achieving up to 28% relative improvement over weakly supervised baselines.
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.