
Data Generation

Data Generation
Dark reactions have the potential to provide the data needed for the next generation of pharmaceutical development.
However, the collection of dark reaction data has been a challenge. There is no incentive for researchers to collect or share this data during current processes.
UnBound Chemicals has developed a proprietary NME recovery system that can recover and digitize valuable materiel from preclinical research activity.
During this recovery process the company generates data that is the same as “dark reactions”.

Dark Reactions
Custom synthesis is a lengthy and manual process. Efforts to improve the process using machine learning has been limited because of the lack of quality data in which to train models.
Nearly all ML models for pharmaceutical development are trained from the open source dataset from the US Patent Office.
This dataset is biased because it only contains successful reactions. For every published patent there are over 20,000 research outcome that go unreported.
These unreported research outcomes are called “dark reactions”.
-
Analyze and digitize material from custom synthesis ( failures, waste, bi-products, side reactions, etc);
-
Recover and return reusable physical material to clients;
-
Build an anonymous database of dark reactions; and
-
Create algorithms to improve the accuracy of custom synthesis planning.