Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models

Konstantin Donhauser, Kristina Ulicna, Gemma Elyse Moran, Aditya Ravuri, Kian Kenyon-Dean, Cian Eastwood, Jason Hartford
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:14321-14343, 2025.

Abstract

Sparse dictionary learning (DL) has emerged as a powerful approach to extract semantically meaningful concepts from the internals of large language models (LLMs) trained mainly in the text domain. In this work, we explore whether DL can extract meaningful concepts from less human-interpretable scientific data, such as vision foundation models trained on cell microscopy images, where limited prior knowledge exists about which high-level concepts should arise. We propose a novel combination of a sparse DL algorithm, Iterative Codebook Feature Learning (ICFL), with a PCA whitening pre-processing step derived from control data. Using this combined approach, we successfully retrieve biologically meaningful concepts, such as cell types and genetic perturbations. Moreover, we demonstrate how our method reveals subtle morphological changes arising from human-interpretable interventions, offering a promising new direction for scientific discovery via mechanistic interpretability in bioimaging.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-donhauser25a, title = {Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models}, author = {Donhauser, Konstantin and Ulicna, Kristina and Moran, Gemma Elyse and Ravuri, Aditya and Kenyon-Dean, Kian and Eastwood, Cian and Hartford, Jason}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {14321--14343}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/donhauser25a/donhauser25a.pdf}, url = {https://proceedings.mlr.press/v267/donhauser25a.html}, abstract = {Sparse dictionary learning (DL) has emerged as a powerful approach to extract semantically meaningful concepts from the internals of large language models (LLMs) trained mainly in the text domain. In this work, we explore whether DL can extract meaningful concepts from less human-interpretable scientific data, such as vision foundation models trained on cell microscopy images, where limited prior knowledge exists about which high-level concepts should arise. We propose a novel combination of a sparse DL algorithm, Iterative Codebook Feature Learning (ICFL), with a PCA whitening pre-processing step derived from control data. Using this combined approach, we successfully retrieve biologically meaningful concepts, such as cell types and genetic perturbations. Moreover, we demonstrate how our method reveals subtle morphological changes arising from human-interpretable interventions, offering a promising new direction for scientific discovery via mechanistic interpretability in bioimaging.} }
Endnote
%0 Conference Paper %T Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models %A Konstantin Donhauser %A Kristina Ulicna %A Gemma Elyse Moran %A Aditya Ravuri %A Kian Kenyon-Dean %A Cian Eastwood %A Jason Hartford %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-donhauser25a %I PMLR %P 14321--14343 %U https://proceedings.mlr.press/v267/donhauser25a.html %V 267 %X Sparse dictionary learning (DL) has emerged as a powerful approach to extract semantically meaningful concepts from the internals of large language models (LLMs) trained mainly in the text domain. In this work, we explore whether DL can extract meaningful concepts from less human-interpretable scientific data, such as vision foundation models trained on cell microscopy images, where limited prior knowledge exists about which high-level concepts should arise. We propose a novel combination of a sparse DL algorithm, Iterative Codebook Feature Learning (ICFL), with a PCA whitening pre-processing step derived from control data. Using this combined approach, we successfully retrieve biologically meaningful concepts, such as cell types and genetic perturbations. Moreover, we demonstrate how our method reveals subtle morphological changes arising from human-interpretable interventions, offering a promising new direction for scientific discovery via mechanistic interpretability in bioimaging.
APA
Donhauser, K., Ulicna, K., Moran, G.E., Ravuri, A., Kenyon-Dean, K., Eastwood, C. & Hartford, J.. (2025). Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:14321-14343 Available from https://proceedings.mlr.press/v267/donhauser25a.html.

Related Material