Healx trains AI to find Covid-19 combination therapies
Healx is using its artificial intelligence platform to develop drug combinations from approved medicines to help treat Covid-19.
The Cambridge biotech company revealed to the Cambridge Independent on Monday that it has trained its algorithms for the virus, and hopes to have candidates available for testing by the “second half of May”.
Its Healnet platform, which integrates biomedical data from sources around the world, will analyse eight million possible pairs of approved therapies and 10.5 billion potential drug triples, based on the 4,000 approved medicines on the market.
Dr David Brown, Healx chairman and co-founder, said: “We have several algorithms that we use and they need different kinds of data and specific training for the disease that we are working on. That’s the phase we have just completed.
“From next week, we shall be working intensively on the predictions – matching existing safe drugs to this specific disease. That phase usually takes three to four weeks because it’s no longer number-crunching – it’s expert drug discoverers working with the output, trying to understand what it means.”
Healx, which won AI Company of the Year at the 2019 Cambridge Independent Science and Technology Awards, is typically focused on using its platform to find therapies for the 95 per cent of the 7,000 rare diseases without treatments.
CEO Dr Tim Guilliams said the rare disease community was particularly vulnerable to Covid-19.
“They don’t usually have treatments available for their disease, and if you add Covid-19 as a comorbidity, they are the ones most at risk,” he told the Cambridge Independent. “This is one of the reasons we are putting significant resources behind this and trying to do it as quickly as possible. It is urgent.”
The knowledge and data surrounding Covid-19 is growing by the day, he added.
“It’s an example of where, unless you learn machine learning and algorithms to keep track of the literature and the dataset, it’s impossible to keep up,” he said.
Dr Brown, the co-inventor of Viagra, said Healx’s complex knowledge graph was like a black box.
“Broadly, you are trying to find a drug that links to a disease, validated by some kind of biochemical pathway and mechanism. It clusters mechanisms and drugs and if you see a big cluster building, that gives you confidence,” he explained.
“Another method uses genetics. We look at how the cell’s DNA expression changes in response to the infection. That kind of data is massive – it requires very heavy computing power but also a lot of human curation of the data going in, because it’s very noisy. It takes us weeks, but we’re finishing that this week.”
The government announced on April 3 that UK hospitals will be involved in the world’s largest Covid-19 clinical trial – one of three it is funding.
In just 15 days, almost 1,000 patients from 132 hospitals were recruited to the randomised RECOVERY trial, which will test drugs including:
- Lopinavir-Ritonavir, commonly used to treat HIV;
- Dexamethasone, a type of steroid used in a range of conditions to reduce inflammation; and
- Hydroxychloroquine, a treatment for malaria.
Dr Brown said the trial, which is led by the University of Oxford, was an illustration of how there were different points to intervene in the infection.
“Hydroxychloroquine and chloroquine act at the very first stage when the virus interacts with a human cell by trying to block that. There are other ways of doing that which we’re looking at as well,” he said.
“Then, once it gets in, the virus gets in and starts to replicate, and that’s what the anti-viral drugs are trying to block, although the clinical data for those doesn’t look too good so far.
“Then you get a cytokine storm, which is where the innate immune system releases lots of biochemicals that cause massive inflammation of the body. Dexamethasone tries to reduce that.
“The reason we talk about drug combinations is that it’s almost certain that one point of intervention is going to be sub-optimal, so hitting in two or three ways is the best thing to do.”
Pre-clinical tests will be carried out once Healx has identified its candidates, before human trials with partner organisations.
Healx is keen to hear from anyone who would like to work with them on the process. The company can be reached by emailing firstname.lastname@example.org.
The company is contacting researchers in Marseille who completed a promising small trial of chloroquine.
“It’s a great example of repurposing and how collaboration can move at things at speed things would never normally move in drug discovery.
“We need to move very first. We need to identify the best drug combinations, get them tested and go to human proof of concept,” said Dr Guilliams.