AI, NLG, and Machine Learning
How AI Could Change Big Pharma
Healthcare AI is already diagnosing patients and analyzing patient data and scans. But can AI also be used to create new medicines?
By Tauren Dyson
March 31, 2020
When a new drug to treat obsessive-compulsive disorder (OCD) entered trials in early February 2020, it made history.
DSP-1181 became the first drug molecule produced using artificial intelligence (AI), shining a ray of optimism on the idea that machine learning can develop drugs more quickly and cheaply than humans can. DSP-1181 was the first non-man made drug to reach Phase 1 clinical trials in under a year, according to AI drug discovery company Exscientia. In fact, it reached that phase five times faster than the normal process for drugs.
“We have seen AI for diagnosing patients and for analyzing patient data and scans, but this is a direct use of AI in the creation of a new medicine,” Andrew Hopkins, founder and chief executive of Exscientia, told BBC.
Exscientia partnered with Japan-based Sumitomo Dainippon Pharma to produce DSP-1181. They programmed the AI method to use a variety of algorithms to create and sort through millions of molecules. The algorithm settled on a group of molecule combinations to test. Then it narrowed down the candidates to a combination of molecules that could attach to the brain receptor that causes OCD.
“There are billions of decisions needed to find the right molecules, and it is a huge decision to precisely engineer a drug,” Hopkins said. “But the beauty of the algorithms is that they are agnostic, so [they] can be applied to any disease.”
The explosion of AI in healthcare has been carried by a wave of investment dollars developing the technology. In 2019, investors poured $4 billion into AI-based healthcare, a number that jumped from $2.7 billion in 2018.
Insilico Medicine is one example of a company that received massive R&D funding for its experimental drug. In less than 50 days, the Hong Kong–based company developed an impressive drug candidate designed to fight fibrosis, netting the company $37 million in 2019.
Other companies, like BenevolentAI, also received a boon from AI healthcare investors. In September 2019, the UK-based firm raised $90 million to develop an AI method to scan for existing drugs that researchers may have overlooked, to fight non-obvious disease targets. That round of funding bumped the company’s valuation to about $1 billion.
Investors hope those dollars can ultimately lower the cost of producing drugs. While the average cost of getting a drug to market has climbed from $1.2 billion to $2 billion over the last decade, the revenue from sales has been cut in half.
Hopkins thinks that AI can help bring down the cost of drug production by 30 percent.
“We have now delivered a few molecules, alongside our partners, into preclinical studies,” Hopkins told NS Business. “Each of those products entered trials after one year—which represents world-class productivity and at a fraction of the cost.”
Pharmaceutical companies are adopting AI to boost drug discovery.
In mid-February 2020, researchers at Massachusetts Institute of Technology (MIT) used AI to attack a dangerous, drug-resistant bacteria. To do this, the AI method needed to understand how to accurately identify the molecular structure of an effective antibiotic.
So a team of scientists at MIT taught an AI algorithm to scour a library of the atomic and molecular structures of close to 2,500 drugs and natural combinations. The algorithm was able to distinguish which molecules could effectively fight the spread of E. coli.
After training the algorithm to accurately spot the molecular structure of effective antibiotics, the researchers needed to teach it which antibiotics to ignore. The team let the algorithm examine more than 6,000 existing compounds designed to fight various human diseases.
The researchers trained the algorithm to disregard existing antibiotic compounds since the bug may have already developed a resistance to the drug. Instead, the algorithm only looked for unique molecular combinations that resembled those that could effectively fight bacteria but didn’t resemble existing antibiotics.
After a few hours, the algorithm pinpointed several potential new antibiotics. One of them, Halicin, which was used to treat diabetes, wiped out a tuberculosis-causing bug known as Mycobacterium tuberculosis, as well as a pesky but harmless gut microbiota called Enterobacteriaceae.
“I think this is one of the more powerful antibiotics that has been discovered to date,” James Collins, a researcher on the team at MIT, told The Guardian. “It has remarkable activity against a broad range of antibiotic-resistant pathogens.”
Medical applications for AI are becoming increasingly attractive to investors. Along with boosting drug discovery and development, AI can offer cost-effective and lower-risk solutions to identify and create new medicines.