AI Model CURE Designs Drugs Based on Transcriptomic Perturbations
A new AI framework, CURE (Cellular Response Engine), formalizes transcriptome-based drug design (TBDD) as a generative inverse problem. The model generates drug molecules conditioned on desired transcriptomic state transitions, addressing the domain gap between biology and chemistry and the sparsity of transcriptomic signals. CURE features a Transcriptome Perturbation Functional Feature Extractor (TFE) that distills function-oriented perturbation embeddings from pre- and post-treatment states. The work is published on arXiv (2605.15243) and represents a computational approach to drug discovery when reliable target structures are unavailable.
Key facts
- CURE (Cellular Response Engine) is a multi-resolution transcriptome-guided diffusion framework.
- The model addresses transcriptome-based drug design (TBDD) as a generative inverse problem.
- It designs drug molecules conditioned on desired transcriptomic state transitions.
- The Transcriptome Perturbation Functional Feature Extractor (TFE) distills perturbation embeddings.
- The approach is useful when reliable target structures are unavailable at scale.
- It handles the domain gap between biology and chemistry.
- The work is published on arXiv with ID 2605.15243.
- The method uses transcriptomic perturbations as a system-level functional readout.
Entities
Institutions
- arXiv