Searching for a drug is a bit like looking for a new house.
Someone might think they have found the perfect home with four double bedrooms, two bathrooms and space for two cars – only to be tempted to compromise on the garage in exchange for another house with a bigger living room or larger kitchen. People hunting for a new house usually realise that they can’t find their ‘dream home’, but will instead pick the best one they can find for the budget they have.
Drug discovery involves a similar process of optimisation and compromise – medicinal chemists must consider lots of different chemical properties when designing a new drug molecule and pick the molecule which best fits their image of a ‘perfect’ drug.
There’s a lot of iterative design, and early on chemists may create chemicals that would never work as drugs – perhaps because they are unstable to metabolism, or because they don’t penetrate and kill cancer cells efficiently enough.
Add in the sheer number of potential molecules that could be prepared and it becomes impossible to create and test every possible molecule. Now, our scientists here at The Institute of Cancer Research in London have developed a method for sifting through many millions of molecules in order to find potential drugs.
Professor Julian Blagg and Dr Nathan Brown, both from the Division of Cancer Therapeutics here at the ICR, are joint authors of a report published in the Journal of Chemical Information and Modeling which describes their workflow for finding and optimising drug molecules. It’s got a clever secret – it uses evolution.
Fragment-based design
The system – known as MOARF (multiobjective automated replacement of fragments) – relies on a process called ‘fragment-based de novo design’, in which a computer ‘cuts’ a target molecule into smaller chemical fragments before replacing them with similar fragments.
A scoring algorithm then ranks these new ‘molecules’ in terms of their potential to act as a drug, feeding this data into another round of chemical changes. The computer gradually ‘focuses in’ on drug-like molecules which begin to ‘evolve’ out of less interesting molecules as the process continues.
Talking about the challenges chemists face in looking for new drugs, Dr Brown said: “The potential number of drug-like molecules is so vast that it’s not only impossible to synthesise them all, but it’s probably also impossible to test them all using computer simulations. We’ve developed a way to computationally scan through thousands of molecules to focus in on those which could be potential drugs – speeding up the process of drug discovery.”
Nicholas Firth, an author of the paper and a PhD student in the in silico Medicinal Chemistry group, told me: “Evolutionary algorithms have been used to design drugs before, but we attempted to balance the requirements of our medicinal chemistry colleagues with an automated design algorithm. Often, a computer can ‘create’ a molecule that is predicted to behave well as a drug, but that is unfeasible to create in the lab due to impossible chemical reactions or cost implications.”
‘Start with an existing drug’
Nicholas told me more about how they worked with medicinal chemists, using computers to sample millions of molecules to find an optimum – the ‘Goldilocks molecule’ which is just right to be an efficient drug.
“We started with a drug structure – Seliciclib,” he said. “This is a new inhibitor of cyclin dependent kinases under clinical investigation by Cyclacel. We knew from the medicinal chemistry team that the central part of the molecule is really important for its ability to bind to its biochemical target, so we kept this scaffold the same and used our algorithm to change the peripheral parts of the molecule.
“There was a huge population of molecules to begin with, and our scoring algorithm picked the ‘fittest’ molecules – those with the most potential to meet project requirements – and we used this to whittle down the selection of potential drug molecules.”
Dr Brown added that this tool can guide chemists towards promising leads: “We are now able to use evolutionary algorithms to design drug molecules using computers. Although we don’t create all of these in the lab, we use the technique to evolve a shortlist of interesting molecules which we then prepare and experimentally test in the lab.”
Just like expanding the search area when looking for somewhere to live, the team at the ICR hope that their new system will allow them to consider and analyse many more potential drug molecules.
And while none of them may be perfect, the most promising could end up as the cancer drugs of the future.
comments powered by