Accelerating novel drug discovery with graph-driven approach

Inspiration

• Imagine if we can reduce the time it takes to develop new drug from 12 years to 12 months
• Imagine if we can reduce the failure rate of drug discovery from 90% to 9%

A graph-driven approach could accelerate the discovery of new drugs and deliver the treatments to patients safer, faster and cheaper.

What it does

1. input a disease that we want to find a novel drug treatment(s)
2. Algorithm will:
3. find list of similar diseases with the input disease
4. find the approved/investigational drugs identified for the list of similar diseases
5. find the highest potential drugs based on their drug targets' association with the input disease
6. list the adverse events of the identified drugs

How we built it

1. Create graph scheme, map data and upload the the data to Tigergraph (data source: https://platform.opentargets.org/downloads/data)
2. Create GSQL query

Challenges we ran into

1. Find and understand the relevant public datasets for drug discovery
2. Prepare / transform the data to Graph structure
3. Understand and identify meaningful relationship among entities to derive hidden insights

Accomplishments that we're proud of

With good quality of data and good graph database technology, it is now possible to rediscover the hidden potential of the same drugs for new treatments to disease.

What we learned

Tigergraph offers SQL-like query language (i.e. GSQL) coupling with graph machine learning libraries. This enables the analytics community who know SQL to quickly pick up GSQL and perform deep link analysis to uncover the hidden insights .

What's next for Discover

To explore the right dosage of drug for disease treatment

Built With

• cosine-similarity
• graph-machine-learning
• gsql
• opentargets
• python
• tigergraph