Interpretable Cell Fate Prediction with Autoencoders

Objectives

• Understand how autoencoders can be used for cell fate prediction

• Learn about adversarial autoencoders and contrastive loss

• Explore latent space interpretation and trajectory analysis

Autoencoder Architecture for Cell Fate Prediction

Discussion

Adversarial Autoencoders

Discussion

Contrastive Loss and Latent Space Structure

Discussion

Latent Trajectory Analysis

Discussion

Keypoints

• Autoencoders can compress image data into interpretable latent representations

• Adversarial training improves reconstruction quality and latent space structure

• Contrastive loss helps separate different cell fate classes in latent space

• Latent trajectory analysis reveals features correlated with treatment outcomes