|
Briefing
|
- Automated Drug Design – MIT researchers developed new computer model that uses machine learning to automate drug design process, resulting to better drugs and faster drug discoveries
- End-to-End Process – Selects lead molecule candidates to base drug on, modifies molecular structure for better potency, and validates chemical results
- AI Training – Model trained on 250,000 molecular graphs from ZINC database, comprised of 3D molecular structures available for public use
- Experiment Results – Initial test resulted in 100% chemically valid molecules, compared to 43% validity for existing systems
- Next Steps – Include shifting work from academic to real pharmaceutical cases, conducting more tests, and assisting human chemists in their work
|
|
Accelerator
|
|
|
Sector
|
Healthcare/Health Sciences
|
|
Organization
|
Massachusetts Institute of Technology
|
|
Source
|
|
|
Original Publication Date
|
July 6, 2018
|
