HCS-DFC: A Diffusion Classifier for Mode of Action Prediction Using Morphological Profiles

Download PDF

Jakub Kościukiewicz, Bartosz Michał Zieliński, Dawid Damian Rymarczyk

Published: 31 Mar 2025, Last Modified: 31 Mar 2025CVDD CVPR2025 Workshop PosterEveryoneRevisionsBibTeXCC BY 4.0
Keywords: HCS, cell painting, conformal prediction, mode of action
TL;DR: We propose diffusion-based classifier for mode of action prediction conditioned on a cell morphology representation
Abstract: Phenotypic-driven drug discovery gains popularity due to the advances in high-content imaging and machine learning, particularly for predicting compound mode of action (MoA) and properties. However, reliance on biochemical assays for label acquisition introduces noise and sparsity, complicating reliability estimation in traditional discriminative models. In this work, we propose High Content Screening Diffusion Classifier (HCS-DFC), reformulating prediction as a conditional generation task to inherently model label distributions and co-dependencies without requiring calibration datasets. By leveraging diffusion models’ ability to capture complex data distributions, HCS-DFC outperforms conformal prediction methods in reliability estimation and achieves state-of-the-art accuracy on both synthetic (MNIST-based multi-task classification) and real-world cell painting datasets.
Submission Type: Original Work
Submission Number: 3