Predicting drug synergy is essential for optimizing combination therapies in cancer treatment. Under extreme data scarcity, existing computational methods struggle to generalize to new cell lines. Although meta-learning approaches have shown promise, a critical limitation lies in their reliance on a unimodal Gaussian prior for task representation. In extreme few-shot settings, this assumption can overly pull the task posterior toward the prior mean, reducing the discriminability of task representations and pushing the model toward a generic mean-value predictor. To overcome this, we propose DiffPro, a novel framework that integrates a Latent Diffusion Model (LDM) as a structural prior. Unlike Gaussian-based methods, our diffusion module learns a flexible, data-driven prior that can model highly complex task distributions. This learned prior better reflects task heterogeneity in few-shot drug synergy prediction under extreme data scarcity, yielding more discriminative task representations. Comprehensive experiments on the DrugComb dataset demonstrate that DiffPro achieves a significant improvement in performance. Notably, in the challenging 5-shot setting, it outperforms the best-performing baseline, delivering approximately a 17.3% relative gain in R2 (0.264 vs. 0.225) and a 4.6% relative reduction in MSE. These results confirm that the diffusion prior successfully regularizes the latent space, mitigating task representation collapse and enabling robust, high-precision predictions for combination therapy design.
Shuting Jin, Xu Guo, Anqi Huang et al.· Proceedings of the 32nd ACM...· 0 citations
Foundation models (FMs) have recently transformed single-cell genomics by learning transferable representations from large-scale single-cell data, enabling a wide range of downstream biomedical applications. Inspired by natural language processing, existing single-cell FMs adapt transformer architectures by treating genes as tokens and cells as sequences. However, transformers are inherently agnostic to input order, while genes lack a natural sequential structure. Current approaches rely on heuristic strategies, such as expression-based gene sorting, to impose positional information, which often fail to capture relative relationships among collectively expressed genes and between gene identities and their expression levels, leading to information loss and limited generalization. In this work, we propose Concord, a novel single-cell FM that explicitly models relationships between gene identities and expression levels through a collaborative rotary attention (CRA) mechanism. Specifically, Concord employs two collaborative attention modes: a gene-to-expression rotational attention that produces gene-enhanced expression representations, and an expression-to-gene rotational attention that yields expression-enhanced gene representations. These two processes provide distinct yet complementary views of the same cell-level expression profile; accordingly, we employ contrastive learning to align their semantic representations. Our theoretical analysis demonstrates that CRA effectively leverages the relative distances in one embedding space to establish stable and consistent dependencies in the other modality. Extensive experiments on various single-cell datasets demonstrate that Concord outperforms existing FMs across various downstream tasks and provides more informative and transferable gene and cell representations. Our code is available at https://github.com/Catchxu/Concord.
Kaichen Xu, Mianpeng Liu, Hao Wu et al.· Proceedings of the 32nd ACM...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
