Machine Learning and its Application in the Development of Novel Cannabinoids

The overall scope of the project seeks to determine if synthetically viable cannabimimetics can be obtained with assistance from datacentric methods. The initial focus is based around designing therapeutics that exert their activity by interaction with the cannabinoid receptors CB1 and CB2. The study may be extended to targeting other proteins which may also be modulated.
The question that this research wishes to answer is can 'Machine Learning techniques be effectively utilised in the development of novel cannabinoid therapeutics?' Furthermore, can these compounds contribute positively to the existing chemical space in medicinal applications.
The overall aim of the project is to therefore adopt Machine Learning methods to generate novel libraries of original ligands that have affinity for proteins already known to be receptive to previously known cannabinoids.
A key parameter in assessing the strength of overall protein-ligand interactions is binding affinity. Thus, libraries need to be screened to assess the strength of overall protein-ligand interactions. To achieve this aim can a reasonable workflow be established, screening a chemical library for binding affinities using a reliable docking assay? Furthermore, can preliminary toxicological and 'drug likeness' assessments be integrated into the workflow.
The project has a strong synthetic aspect and as such is also expected to present some challenging target molecules. This itself could in part shape the direction that the project takes, for example, the exploration and expansion of a particular synthetic method.