Old drugs, new uses

Drug development is an expensive, resource-heavy and slow business. It typically takes more than a decade and costs more than £1bn for a drug to make it from the lab to patients. But of drugs that begin preclinical testing, only five in 5,000 ever make it to clinical trials; and ultimately only one of those makes it to market. That means that decades' worth of work and hundreds of millions of pounds are locked up in many thousands of drugs that no one has any use for. That is until now.

Drawing on open-source software developed at Harvard and MIT, a Salt Lake City startup is using image recognition and AI to find new uses for old drugs. By feeding software thousands of images of diseased cells, Recursion Pharmaceuticals works to identify whether any of the more than 2,000 drug compounds they are testing will make the sick cells more healthy.

The company can identify potential treatments right away without delving into the underlying mechanisms of particular diseases, making the discovery process much quicker. The company was founded in 2013 by researchers from the University of Utah, and has grown to a team of 40. "In some sense we are standing on shoulders of giants," says co-founder and CEO Chris Gibson, referring to the use of open-source software. CellProfiler was built as a free, general purpose tool for biologists to measure and analyse cells; its developers have welcomed Recursion's use of it for drug development.

Recursion uses CellProfiler to measure as many features as possible in a cell. On average they measure about 800 features and observe the shape and morphology of a disease-affected cell and how it differs from a healthy one. The idea is to find drugs that can fix a diseased cell and reform it to health. Currently, the company can test up to 1,000 drugs against two disease models per week. "It's facial recognition for cells," explains Tim Considine, Senior Vice President of Strategic Development at Recursion.

The approach is a boon for the treatment of rare diseases. As many as 36 million people in the EU alone live with a rare disease – defined in the EU as one that affects fewer than one in 2,000 people – but most have no treatment options at all. Rare disease drug development gets little to no funding as there is little financial incentive for major drug companies to address them. Recursion's method increases the scale at which drugs are discovered, potentially saving years of development time and billions of dollars. The company hopes to find treatments for as many as 100 rare diseases in the next 10 years.

Its early successes are promising: it took the company less than a year to identify more than 12 compounds across seven disease states that show sufficient potential to advance to preclinical development. So far, Recursion has already identified 14 potential treatments for rare diseases.

There are around 7,000 different types of rare diseases and disorders. Cerebral cavernous malformation (CCM) is one of them, affecting 65,000 people a year in the US alone. It is caused by abnormal blood vessels that form raspberry-like clusters in the brain and spinal cord. If these angiomas bleed or press against structures in the central nervous system, they can produce seizures, neurological deficits, haemorrhages and headaches. The only treatment available for it is invasive brain surgery, which is a high-risk procedure. Recursion identified a drug that could be used for treatment. Clinical trials are set to start early 2018.

Recursion's efforts have attracted the attention of some industry giants. Currently, three of the top ten pharma companies have opened up their libraries of unused drugs in the hope that Recursion will find new uses for them. Further, the National Institutes of Health (NIH) has given a $2m grant to conduct a study on senescence (ageing). It is also collaborating with Recursion to help find treatments for undiagnosed rare diseases, which generally affect fewer than 10 people at a time. Giving hope to some patients whose doctors are at a loss, unable to define their condition.