Our Research
The Romano Lab conducts original research at the interface of clinical/translational informatics and environmental health, with particular focus on developing new artificial intelligence (AI) models that explain the biology underlying environmental risk factors and mechanisms of disease. Other areas of research interest include autoimmune-related adverse events, maternal environmental health, and natural products drug discovery. Dr. Romano leads the development of several major biomedical knowledge bases, including ComptoxAI, VenomKB, and the Alzheimer's Knowledge Base.
Heterogeneous Graph Machine Learning
Geometric machine learning - or graph machine learning (GraphML) - is a branch of machine learning that discovers patterns in graph-formatted (network) data structures, and uses those patterns to make intelligent decisions on future data.
Computational Toxicology
Computational toxicology is a field that uses computational models to predict and explain the toxic effects of specific chemical compounds on humans and ecosystems. For decades, this has been done almost entirely using a small set of techniques (like quantitative structure-activity relationship (QSAR) modeling), but recent advances in machine learning and knowledge graph construction have opened up new possibilities. We are particularly interested in using structured biomedical knowledge (e.g., Adverse Outcome Pathways) and emerging AI methods (e.g., graph neural networks) to improve the accuracy and explainability of predictive toxicology.
Funding
Our research is supported by:
- NSF: SCH: An intraoral device for multimodal data integration and risk modification connects the oral microbiome to systemic health (2025-2029)
- NIH/NLM: Discovering clinical endpoints of toxicity via graph machine learning and semantic data analysis (2021-2025)