Integrating computational chemistry models of ion channels to build a holistic cardiac safety prediction tool for drug discovery

Ensuring that any potential new medicine does not cause harmful arrhythmias has long been a great challenge. The major cause of this is the ability of many compounds to interfere with ion channels that are crucial to the heart's regular beat. In this project computational chemistry tools will be combined with cutting edge machine learning approaches to build a holistic model for cardiac safety that will be of great benefit to drug discovery. The aspiration is that a scientist could draw a chemical structure that would then be placed inside each of the most important cardiac ion channels. Quantum mechanical calculations would provide the binding energy between the molecule and each of the ion channels. Finally, the computed binding energies with all of the key ion channels would be integrated to provide a prediction of the in vivo activity of the compound. The current alternative to this approach is to test the compound using animals; not only could the computational approach avoid many of these animal tests, it would be superior to them because it provides an explanation for any problems and can be used to design alternative, safe molecules. The multidisciplinary team that will be supporting the project includes computational and synthetic chemists, biologists and even clinicians working in the adjacent hospitals providing cardiovascular care. The project will see the student first build a few models of individual ion channels and validate these with all of the known compound binding data. They will then select a set of compounds with known cardiac safety and compute their binding energy in all of their own computational models as well as all others available in the group at that point in time. This will provide a computed profile of the compounds for many ion channels. The link between these profiles and the known in vivo activity of the compounds will be explored using the latest machine learning techniques. These techniques are currently of high interest across the pharmaceutical and other industrial sectors. Subsequently, the student would design the last part of the project themselves with possibilities including making compounds to test their own predictions, testing compounds in in vitro assays or applying the machine learning to other therapeutic areas, such as neuroscience, that also have links to ion channel activity. This project is tailored to address two of the key MRC themes, namely Quantitative Skills and Interdisciplinary Skills. Within the quantitative skills, the project will train the student in ALL of the listed areas (mathematics, statistics, computation, data analytics and informatics, machine learning and Artificial Intelligence, developing digital and technology excellence) in this case, the domain of application will be to whole organism and whole tissue findings deriving from in vivo experiments (including clinical findings) and to a range of in vitro experimental findings. The project team and project outline reflect the high interdisciplinarity of the project that will require the student to work at the interface where chemical, physical and computational science meet in vivo biology and clinical application. The computational methods involved will see the underlying physics of molecular interactions described accurately in a way that will provide chemical insight. The student will then apply the machine learning techniques to forge a link from the molecular level to the whole organism/tissue level.