Original article was published on Artificial Intelligence on Medium
Japanese startup Elix is using AI to research COVID-19 drugs
Neural networks greatly speed up the hunt for effective medications
The global search for a COVID-19 vaccine and treatment is on. Researchers are using every resource at their disposal to try and find a breakthrough against the killer virus. That includes artificial intelligence (AI).
The tech has previously shown promise as a tool for drug development, but now researchers are putting it to the ultimate test — finding a drug that treats COVID-19 in record time. A Japanese startup called Elix is leading the charge thanks to deep learning software from Nvidia.
Needle in the Haystack
Finding drug molecules to treat diseases used to be a task that took years. Now, sophisticated AI systems can significantly condense the process. Even so, the number of potential drug-like molecules out there is somewhere in the neighborhood of 1060.
Shinya Yuki, co-founder and CEO of Tokyo-based startup Elix, puts it in perspective by saying, “Even if you hypothetically checked one molecule per second, it would take longer than the age of the universe to explore the entire chemical space.”
He adds, “AI can efficiently explore huge search spaces to solve difficult problems, whether in drug discovery, materials development or a game like Go.”
The game that Yuki references became popular in the AI space last year after Google’s AlphaGo algorithm forced Korean Go master Lee Se-dol into retirement, citing the fact that AI is unbeatable. Go has been around for more than 2,500 years and features a virtually unlimited number of moves.
In a sense, it’s a lot like looking for one helpful drug molecule in an endless pile of useless ones. That’s why Yuki’s company is using AI to accelerate the process.
Elix uses deep learning to build neural networks that predict the properties of molecules far faster than traditional computer simulations. To speed up the process, the researchers are using molecules that are either in a clinical trial or have FDA approval. They want to find potential candidates that could be repurposed to treat the coronavirus.
“Speed is critical, and drug-repurposing can help identify candidates with an existing clinical safety record, significantly reducing the time and cost of drug development,” says Yuki.
Elix’s efforts are starting to pay off. The startup recently released a paper detailing the various drug molecules that its AI model flagged as potential COVID-19 treatments. Notably, an antiviral drug called remdesivir is on the list.
The same drug recently gained FDA emergency use authorization to treat critically-ill coronavirus patients. Although remdesivir doesn’t cure COVID-19, clinical trials showed that it can decrease the time to recovery in some patients.
Per the FDA, “The emergency use authorization allows for remdesivir to be distributed in the U.S. and administered intravenously by health care providers, as appropriate, to treat suspected or laboratory-confirmed COVID-19 in adults and children hospitalized with severe disease.”
The fact that remdesivir is on the list generated by Elix’s AI is good news. For one, it somewhat confirms the results of earlier clinical trials and suggests that the drug might be effective. More importantly, however, it proves that the algorithm is on the right track. Since it flagged remdesivir, which is showing promising results, other drugs on the list might be effective as well.
Cutting Edge Tech
The process which Elix uses to narrow down its drug selection is complex. It includes several layers of neural networks and predictive AI models. The first layer analyzes a database of drug molecules and determines which ones have the physical or chemical properties to treat COVID-19.
Meanwhile, generative models start from scratch to create new molecular structures that might be promising. The issue with this, however, is that not all of the structures are found in nature. As such, a third neural network is needed. It is a retrosynthesis model that helps researchers determine if it’s even possible to synthesize the generated molecules.
To accomplish this feat, Yuki and his team utilize multiple Nvidia DGX Station systems. These GPU-powered AI workstations allow neural networks to train up to six times faster than using a single GPU. That acceleration is vital to making the generative models a practical part of the research. Otherwise, it would take weeks for them to reach the lowest error rate possible.
Each DGX station features four of Nvidia’s V100 Tensor Core GPUs. This allows the Elix team to run multiple experiments simultaneously and use larger AI models. It’s worth noting that Elix is a member of Nvidia Inception, a program that helps startups reach their market faster.
Although it is currently focusing on identifying COVID-19 drugs, Elix isn’t shy about using AI. The startup also harnesses the technology’s power across several other applications, including molecular design for material informatics. Its client list includes noteworthy companies like Bridgestone and Riken.
Elix used a similar process to develop a new molecule for the global chemical company Nippon Shokubai. Together, the duo created a molecule that helps blend ink without posing a high risk for skin irritation.
Meanwhile, Elix also develops computer vision models for self-driving car applications.
Just the Beginning
As mentioned, drug research is an extremely inefficient task when done manually. That makes it a golden application for artificial intelligence. This isn’t the first time that it has been used.
Last July, an Australian AI system developed the first computer-generated human drug. It was an add-on that aims to boost the efficiency of traditional flu vaccines.
Meanwhile, a newly developed AI drug known as DSP-1181 entered the human trial phase earlier this year. The drug, discovered by Exscientia, is being developed to treat obsessive-compulsive disorder (OCD). At the time, the team behind the drug said, “This year was the first to have an AI-designed drug but by the end of the decade all new drugs could potentially be created by AI.”
Thanks to AI-focused applications like Elix’s, humanity could be closer than ever to finding a viable treatment for COVID-19. With little time to waste, artificial intelligence is a valuable ally in the fight against the virus.