scTrilemma: Balancing Identity, Invariance, and Fidelity in Single-Cell Representation Learning

2026-09-29

AI Summary

This study addresses the trilemma in single-cell representation learning, where identity recognition, invariance, and fidelity are difficult to achieve simultaneously. We propose scTrilemma, a variational autoencoder (VAE) equipped with a latent bottleneck. This method routes gene variation into embedding, decoder, or prior spaces via a gating mechanism and conditions the prior on pseudo-batch contexts, enabling effective disentanglement of biological signals from confounders without requiring labels or auxiliary losses. Zero-shot evaluations demonstrate that scTrilemma concurrently achieves state-of-the-art performance across all three metrics on the CZ CELLxGENE dataset while faithfully preserving biological states and pathway structures.

Abstract

Single-cell RNA-seq representation learning is fundamentally label-free: cell identities, states, and contexts are not fixed training targets, so what constitutes signal or nuisance is analysis-dependent. A single representation must therefore preserve biological identity and state, remain robust to nuisance context, and retain the gene-level variation needed for expression analysis, three demands we call the representation trilemma. To tackle this problem, we introduce scTrilemma, a latent-bottleneck VAE that routes expression-derived variation to the embedding, the decoder, or the prior rather than forcing all of it through one embedding. It gates gene tokens by expression, routes the cell representation through the decoder, and conditions the prior on unlabeled pseudo-bulk context, under a single reconstruction objective and without target annotations or auxiliary representation losses. In release-based zero-shot evaluation on successive CZ CELLxGENE Census releases, scTrilemma leads all three demands at once and preserves biological-state, differential-expression, and pathway structure across multiple disease settings. Latent interventions further show that context can be removed at almost no cost to the other demands, leaving identity against fidelity as the remaining tension. Code is publicly available at https://github.com/yunhak0/scTrilemma.

Problem

• Single-cell representation learning
• Representation trilemma
• scRNA-seq
• Latent space
• Zero-shot evaluation

Innovation

• single-cell representation learning
• latent-bottleneck VAE
• representation trilemma
• gene token gating
• zero-shot evaluation