Researchers at Harvard Medical School have developed an AI model that could change the way new drugs are discovered.
The tool, called PDGrapher, identifies cellular changes that can reverse disease states. It is freely available and described in Nature Biomedical Engineering.
Traditional Drug Discovery
Drug discovery has long been a slow and uncertain process. Traditional methods focus on testing one protein or one compound at a time.
Success often comes only after years of trial and error; however, PDGrapher offers a different approach.
Instead of isolating one target, it analyzes many drivers of disease. It then predicts which genes or drug combinations can restore diseased cells to a healthy state.
The model does not guess blindly. It makes informed decisions based on cellular networks.
The Workings
PDGrapher uses a type of AI called a graph neural network. This system examines connections between genes, proteins, and pathways.
It simulates what happens if certain cellular processes are switched on or off. From there, it identifies the treatments most likely to reverse disease.
This method saves time and also reduces the need for endless testing of compounds. Instead, it directs researchers toward the most promising options.
Also read: Assort Health Raises $50 Million for AI Healthcare
Evidence
The team trained PDGrapher on cell data before and after treatment. They then tested it on 19 datasets from 11 cancer types. The tool performed with notable accuracy.
- It confirmed known drug targets that had been excluded from training.
- PDGrapher identified KDR (VEGFR2) as a target in non-small cell lung cancer, in line with clinical findings.
- It flagged TOP2A, already targeted by chemotherapy drugs, as a promising option for slowing tumor spread.
- In head-to-head comparisons, PDGrapher ranked correct targets up to 35 percent higher than similar models. It also delivered results up to 25 times faster.
Importance
Complex diseases such as cancer often resist treatments that attack a single pathway. Tumors adapt and develop resistance.
PDGrapher addresses this challenge by revealing multiple targets at once. The model could also help scientists study diseases that remain poorly understood.
Researchers are already applying it to brain disorders, including Parkinson’s and Alzheimer’s disease.
Collaborations with Massachusetts General Hospital are using it to investigate X-linked Dystonia-Parkinsonism, a rare genetic condition.
The potential uses go beyond discovery; PDGrapher could help design personalized treatments.
It could analyze a patient’s cell profile and guide doctors toward drug combinations tailored to individual needs.
PDGrapher may also provide researchers with deeper knowledge. It can explain why certain treatments succeed by identifying the biological drivers behind effective therapies.Â
This insight may lead to new breakthroughs across medicine. As study author Marinka Zitnik explained, traditional methods resemble tasting dish after dish to find one with the right flavor.
PDGrapher, in contrast, shows how to select and combine ingredients with precision. Her team hopes the model will create a roadmap for reversing disease at the cellular level.
If successful, it could mark a turning point in biomedical research and drug development.
