Resolving superposition in AI for interpretability and cross-modal alignment in patient-neuronal images

2026-06-30

AI Summary

This study addresses the challenge of concept superposition in neural networks caused by dimensional bottlenecks in high-dimensional biological imaging, which distorts latent space geometry and compromises interpretability. Applying sparse autoencoders (SAEs) to over 100,000 multiplexed images of Parkinson’s disease and healthy neurons, this work provides the first evidence of how superposition contaminates geometric structure and introduces an SAE-based geometric purification strategy. Furthermore, by integrating Gromov–Wasserstein optimal transport with single-cell state modeling, the authors propose GW-map—a novel, reference-free method that enables de novo cross-modal alignment between imaging data and scRNA-seq without requiring spatial transcriptomics as a scaffold. The framework successfully reconstructs neuropathological pathways such as the calcium–AIS scaffold, establishing a high-fidelity, interpretable foundation for scalable spatial biology.

Abstract

Artificial intelligence is transforming our capability to solve biological challenges. In dimensionality bottleneck regimes exacerbated by high-dimensional biological data, Neural networks force distinct concepts into the lower dimensions known as superposition. Although this superposition is widely known to hinder interpretability, its impact on corrupting the geometry of latent spaces remains critically overlooked. Here, we utilized sparse autoencoders (SAEs) trained on over 100,000 multiplexed images of patient-derived Parkinson's disease and healthy neurons to resolve superposition. This approach bypasses the mathematical non-uniqueness of feature attribution by shifting to interpretable latent representation analysis. We theoretically and empirically demonstrate that superposition contaminates representational metric spaces, and thereby SAEs successfully recover geometric fidelity. By treating these geometrically purified representations as single-cell state vectors, we adapted single-cell RNA sequencing (scRNA-seq) data analysis methodologies directly to the image domain. Finally, we introduce GW-map, utilizing Gromov-Wasserstein optimal transport to align these image representations with authentic scRNA-seq data \emph{de novo}. This coupling reconstructs hierarchical neuronal pathology pathways such as Calcium-AIS scaffold, without reference spatial transcriptomics, establishing a scalable foundation for spatial biology. Code is available at https://github.com/jijihihi/Bio_superposition

Problem

• superposition
• interpretability
• cross-modal alignment
• latent space geometry
• high-dimensional biological data

Innovation

• superposition resolution
• sparse autoencoders
• geometric fidelity
• cross-modal alignment
• Gromov-Wasserstein optimal transport