An ‘automated design’ system for new drugs could help develop the complex therapies needed for many medical conditions while also improving drug safety and efficiency, new research from the University of Dundee has revealed.
Researchers from the College of Life Sciences at Dundee, in collaboration with partners in North America, have shown that an automated computational process analysing huge amounts of existing data could provide a valuable new tool in drug discovery.
Their ‘Moneyball’ approach mimics the creative process of human chemists, where drug molecules are steadily improved through successive cycles of design and selection.
‘One of the things that makes drug discovery so hard is that you’re trying to improve several different properties at the same time,’ said Professor Andrew Hopkins, chair of Medicinal Informatics at Dundee. ‘Evolution is a mechanism that can be applied to solving these kinds of optimisation problems, and the iterative process of adaption and selection of hundreds of thousands of possible solutions can be simulated in a computer.
‘We have effectively proved the concept of automated design of new compounds, showing that by using algorithms to process massive amounts of data we can tackle problems of huge complexity. The system solves the design problem by using computational evolution to mimic the design process of human chemists but running it on a very large scale.’
We have effectively proved the concept of automated design of new compounds
Professor Hopkins and colleagues initially used an automated adaptive design approach to look at Donepezil, an existing drug used to treat Alzheimer’s disease.
‘We took the structure of Donepezil as a starting point and from there the system evolved its structure, computationally, over many generations to a variety of different profiles across a range of drug targets. The predicted profiles were then tested experimentally and we found that 75% of them were confirmed to be correct.
‘This proof of concept shows that we could make significant advances in discovering and designing complex drugs, which could lead to improvements in safety and efficacy, while also potentially reducing the cost of drug discovery, which is a high-risk and expensive process.
‘Just a few years ago this would not have been possible because we need the existing drug data to build on and it was not held in a way that it could be analysed like this. But there have been significant developments, aided by groups such as ChEMBL in Cambridge, who are funded by the Wellcome Trust, in making drug design data available in a format computers can process. What we have found particularly exciting is the way the algorithm has been able to learn from the human experience of drug design and mimic it on a massive scale to solve complex design problems.’
A new spin out company, Ex Scientia, has been formed to commercialise the technology.
The research is published in the journal Nature and funded by the Biotechnology and Biological Sciences Research Council.