recent years, however, the pipeline for novel antibiotics has begun to fill again, thanks mainly to small, innovative biotech companies and their academic collaborators. And among these, computational and structure-based approaches are yielding dividends.

Overcoming the challenges of resistant bacteria For almost a century, the city of Oxford has played an important role in the stories of both antibiotic development and computer-aided drug design. Nobel- winning pathologist Howard Florey, who was involved in developing penicillin, held a chair in pathology there; it was there that the first patient to be given penicillin was treated, and it was there that Dorothy Hodgkin deciphered the three- dimensional structure of the penicillin molecule. She was also a pioneer in structural biology, one of the disciplines has that paved the way for today’s

”Our methods can rapidly select groups of molecules with different chemical structures that are likely to have similar properties and bind a target in a similar way”

bacterial cell wall: a complex chemical structure that is unlike anything in the human body. Countless millions of lives have been saved by these drugs since they were first introduced in the 1940s, but resistance to them has been around almost as long. So-called Gram-negative bacteria, which have a double-layered cell wall that many antibiotics are unable to cross, are particularly difficult to target with antibiotics. These pathogens predominate on the internationally maintained lists of ‘priority organisms’ for which new drugs are most acutely needed. No truly novel drug to treat infections

with Gram-negative bacteria has entered the clinic for over half a century. In | @scwmagazine

computational methods in drug discovery. The first definitive success story of such methods was the design of HIV protease inhibitors in the late 1980s. Move on another decade, and algorithms for simulating the interaction between a putative drug (generally a smallish, organic molecule) and its protein target had become fast and reliable and, at least in their basic application, required very little user input. In the late 90s, Graham Richards, head of chemistry at Oxford and a successful biotech entrepreneur, developed the so-called ‘screensaver lifesaver’ project to exploit millions of personal computers’ idle time in the service of drug design, screening a vast library of molecules against proteins known to be involved in cancer. This technology was spun out into a drug discovery consultancy, Inhibox; in 2017 the company was re-launched with a changed focus as Oxford Drug Design. ‘The methods we have developed can be applied to any therapeutic area, but we chose to focus on antibiotic resistance because of the immediacy and severity of the unmet medical need’, says Richards.

Richards and his colleagues have built up a database of, currently, more than 100 million molecules that can be synthesised easily and have suitable physical and chemical properties to be used as drugs. In parallel, they have developed extremely fast, robust algorithms for screening these compounds against any target protein with a known three-dimensional structure. ‘There’s no point in setting up an enormous database if it takes months to search’, says the company’s CEO, Paul Finn. The search algorithms employ a novel ‘molecular descriptor’ in which the shape and charge distribution of each molecule is represented as a string of real numbers. ‘Our methods can rapidly select groups of molecules with different chemical structures that are likely to have similar properties and bind a target in a similar way’, adds Finn. The first antibiotic discovery project to benefit from these methods is focusing on an enzyme that is critically important for bacterial protein synthesis but that is sufficiently different from the equivalent human enzyme to be safely inhibited by drugs. The colleagues screened their database against leucyl tRNA-synthetase from Escherichia coli, an often pathogenic Gram-negative bacterium. ‘We identified some likely compounds and asked collaborators in Latvia to synthesise them, and they proved to be active against the bugs’, says Finn. ‘The next step is to tweak these structures, so they stay in the bacterial cells for long enough to be truly effective’. Many potent compounds fail as drugs

because they cannot cross the Gram- negative cell wall or are rapidly ejected from the bacteria. The company is using machine learning techniques, including neural networks and support vector machines, to develop rules for predicting which molecules will be substrates for bacterial efflux pumps, and therefore ejected from the cells. They hope to begin clinical trials within a few years, initially focusing on urinary tract infections with resistant, Gram-negative bacteria. Across the pond, Forge Therapeutics in San Diego, USA, has developed a drug design platform to target enzymes that depend on metal ions in their reaction mechanisms. Forge was spun out of the University of California San Diego in 2015, with technology based on the work of Seth Cohen, a bio-inorganic chemist. ‘What makes our computational

platform unique is the specific focus on understanding and simulating the properties of the metal and its interactions with the protein target and candidate drug molecules’, says Forge’s COO, David

August/Septemebr 2018 Scientific Computing World 19


Kateryna Kon/ WhiteMocca/

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36