MoLF: Mixture-of-Latent-Flow for Pan-Cancer Spatial Gene Expression Prediction from Histology

TL;DR

MoLF introduces a pan-cancer, histology-to-spatial gene expression predictor by coupling a Transformer-based VAE that learns a latent gene manifold with a conditional flow-matching model whose velocity field is expressed via a Mixture-of-Experts. The MoE routing decouples heterogeneous tissue signals, enabling efficient multimodal generation and robust cross-cancer generalization, including zero-shot cross-species transfer. Empirical results on the HEST-1k pan-cancer benchmark show state-of-the-art performance across highly variable genes and Hallmark pathways, with ablations confirming the importance of the MoE architecture and spatial encoding. The work demonstrates that decomposing histology into reusable morphological primitives via specialized experts yields improved generative alignment and generalization in histogenomics.

Abstract

Inferring spatial transcriptomics (ST) from histology enables scalable histogenomic profiling, yet current methods are largely restricted to single-tissue models. This fragmentation fails to leverage biological principles shared across cancer types and hinders application to data-scarce scenarios. While pan-cancer training offers a solution, the resulting heterogeneity challenges monolithic architectures. To bridge this gap, we introduce MoLF (Mixture-of-Latent-Flow), a generative model for pan-cancer histogenomic prediction. MoLF leverages a conditional Flow Matching objective to map noise to the gene latent manifold, parameterized by a Mixture-of-Experts (MoE) velocity field. By dynamically routing inputs to specialized sub-networks, this architecture effectively decouples the optimization of diverse tissue patterns. Our experiments demonstrate that MoLF establishes a new state-of-the-art, consistently outperforming both specialized and foundation model baselines on pan-cancer benchmarks. Furthermore, MoLF exhibits zero-shot generalization to cross-species data, suggesting it captures fundamental, conserved histo-molecular mechanisms.

Figures (8)

• Figure 1: Overview of MoLF Architecture.Stage I: A Transformer-based VAE compresses high-dimensional gene expression into a structured latent manifold. Stage II: A Conditional Flow Matching model with Mixture-of-Experts (MoE) velocity estimation learns to transport noise to this latent manifold, conditioned on H&E morphological features.
• Figure 2: Qualitative Results of Synthetic Gaussian. By routing conditions to specialized experts, the MoE architecture avoids the averaging artifacts seen in the dense baseline.
• Figure 3: Macro-scale generative transport.Left: Initial state ($t=0$) showing Gaussian noise projected in UMAP space (red). Right: After one ODE step, the noise maps to a structured manifold (red) that aligns with the sample's ground truth distribution (blue), confirming global distributional matching.
• Figure 4: Micro-scale trajectory analysis. Trajectories for six random patches from a single sample. The model transports a random starting noise vector (green dot) to a predicted latent (red 'X') via a single straight-line step. The prediction lands in close proximity to the specific ground truth destination (cyan star), demonstrating accurate conditional coupling.
• Figure 5: Expert specialization analysis.Left: Gene latent embeddings colored by cancer type. Right: The same gene latent embeddings overlaid with activated colored experts. The lack of distinct clustering of experts in the right panel indicates that experts are not segregated by cancer type, but rather contribute collaboratively across the dataset.