Computer-aided drug discovery
Based on the pioneering work at Oxford University of Prof. Graham Richards, Oxford Drug Design has spent many years developing a diverse, proprietary toolkit of powerful virtual screening, ligand-based and structure-based drug design technologies.
We apply our custom-developed AI/ML tools specifically to each drug discovery effort rather than trying to solve every problem in the same way.
This leads to superior insights and the acceleration of our discovery outcomes.
Innovative chemical representations
To be useful, any AI drug discovery method must be based on a meaningful representation of chemical structure.
A key differentiating strength of Oxford Drug Design is the special emphasis placed on generating such chemical representations.
ElectroShape (EShape)
Our proprietary tool EShape captures the complexities of molecular shape and charge in a compact and easily searchable form.
Incorporated into our molecular databases, molecular similarity comparisons can be then assessed at speeds of one million/sec. This enables data mining of very large scale drug-relevant chemical spaces.
The results critically include virtual compounds that are readily synthesizable from accessible chemical building blocks and sustain robust and reliable chemical reactions.
Topological data analysis (TDA)
TDA is the field of mathematics providing a set of topological and geometrical tools to analyse complex, high-dimensional datasets. We have incorporated TDA into our molecular shape/fit AI methods.
Oxford Drug Design is applying and developing innovative analyses of molecules via TDA to extract persistent (homological) features from their three-dimensional shapes.
These novel representations then enrich and complement existing ones, thus in turn generating important new opportunities for our machine learning applications.
AIScape
AIScape integrates our machine learning models into a powerful visualization and data analysis tool.
An important aspect of these models is their interpretability, providing our scientists with an understanding of the reasoning behind the model prediction. This is preferable to “black box” AI methods lacking such essential understanding.
AIScape enables our domain scientists to gain superior insights into the relationships between chemical structures and their desired biological activity.