Contrastive learning of cell state dynamics in response to perturbations

2024-10-15

arXiv.org

AI Summary

This study addresses the challenges of insufficient annotations and poor generalizability in fine-grained dynamic representation learning for time-lapse live-cell imaging. We propose DynaCLR—the first self-supervised contrastive learning framework tailored for single-cell time-series images. DynaCLR integrates single-cell trajectory tracking with time-aware contrastive learning, enforcing temporal proximity of embeddings for neighboring timepoints via a time-regularized loss, thereby enabling unsupervised modeling of cellular states and organelle dynamics. Crucially, it explicitly incorporates temporal structure into contrastive learning, supporting cross-experiment generalization and diverse downstream tasks: viral infection classification accuracy exceeds 95%; transient mitotic events and virus-induced organelle responses are reliably identified; asynchronous cellular responses are aligned; and cross-modal state distillation is achieved. The framework incorporates temporal data augmentation, end-to-end training, and a napari visualization plugin.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Representation Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

We introduce DynaCLR, a self-supervised framework for modeling cell dynamics via contrastive learning of representations of time-lapse datasets. Live cell imaging of cells and organelles is widely used to analyze cellular responses to perturbations. Human annotation of dynamic cell states captured by time-lapse perturbation datasets is laborious and prone to bias. DynaCLR integrates single-cell tracking with time-aware contrastive learning to map images of cells at neighboring time points to neighboring embeddings. Mapping the morphological dynamics of cells to a temporally regularized embedding space makes the annotation, classification, clustering, or interpretation of the cell states more quantitative and efficient. We illustrate the features and applications of DynaCLR with the following experiments: analyzing the kinetics of viral infection in human cells, detecting transient changes in cell morphology due to cell division, and mapping the dynamics of organelles due to viral infection. Models trained with DynaCLR consistently achieve $>95%$ accuracy for infection state classification, enable the detection of transient cell states and reliably embed unseen experiments. DynaCLR provides a flexible framework for comparative analysis of cell state dynamics due to perturbations, such as infection, gene knockouts, and drugs. We provide PyTorch-based implementations of the model training and inference pipeline (https://github.com/mehta-lab/viscy) and a user interface (https://github.com/czbiohub-sf/napari-iohub) for the visualization and annotation of trajectories of cells in the real space and the embedding space.

Problem

• Learns robust representations of cell dynamics from time-lapse images
• Enables diverse downstream tasks with minimal human annotations
• Generalizes to both in-distribution and out-of-distribution datasets

Innovation

• Self-supervised contrastive learning for cell dynamics
• Temporal regularization with single-cell tracking
• Generalizable embeddings for diverse biological tasks