AROMA: AI Framework for Virtual Cell Genetic Perturbation Modeling
A team of researchers has introduced AROMA, an augmented reasoning framework designed for modeling genetic perturbations in virtual cells. This framework combines textual data, graph-topology insights, and protein sequence characteristics to forecast alterations in molecular states due to genetic changes. AROMA employs a two-phase optimization approach to ensure precise and understandable predictions. Additionally, the researchers developed two knowledge graphs and a dataset for perturbation reasoning, named PerturbReason, which includes more than 498,000 samples. Experimental results indicate that AROMA surpasses current methodologies.
Key facts
- AROMA stands for Augmented Reasoning Over a Multimodal Architecture.
- It models perturbation-target dependencies using textual, graph, and sequence data.
- Two-stage optimization ensures accuracy and interpretability.
- PerturbReason dataset includes more than 498,000 samples.
- Two knowledge graphs were built as reusable resources.
- AROMA outperforms existing methods in experiments.
- The work is published on arXiv with ID 2604.20263.
- Virtual cell modeling is essential for biological mechanism studies.
Entities
Institutions
- arXiv