Artificial Intelligence ‘Boot Camp’ Aims to Accelerate Drug Discovery 
Machine learning is playing an increasingly important role in computing and artificial intelligence.

TEL AVIV – Ninety percent of drug candidates fail in clinical trials because of unexpected safety issues or lack of efficacy in human subjects, according to Noga Yerushalmi, Investment Director at M Ventures, the strategic, corporate venture capital arm of Merck.

There have been dramatic improvements in omics technologies – that is the collective technologies used to explore molecular behaviour, such as genomics, proteomics, metabolomics, metagenomics and transcriptomics.  And this is enabling scientists to analyze potential drug targets more efficiently in terms of their potential impact on a pathogen or disease. 

But there is no automated solution that harnesses all preclinical data in a way that allows for a reliable assessment of the clinical trial readiness of a new drug candidate.

Now, some pharma and research initiatives are hoping to change that, among them AION Labs.  Together with its German partner BioMed X, the Israeli-based research alliance, hosted 15 teams of computational biologists, AI researchers and biomedical scientists for a ‘boot camp’ that aims to speed up the drug discovery process and lower costs – harnessing AI technologies in novel ways. 

Mission: identify critical safety issues before expensive clinical trial stage 

The latest boot camp cohort

The teams were charged with making a proposal for the development of a versatile, next-generation computational platform that can identify hidden safety liability and lack of efficacy, and close identified gaps in the drug candidate pre-clinical data package. 

“The big problem in drug discovery and development is that after a lot of experiments have been done and we have a new drug candidate, many of these candidates fail in very expensive clinical trials in humans,” Dr. Christian Tidona, founder and managing director of the

BioMed X Institute told Health Policy Watch. “That’s because humans are different from mice or a dish in the lab.”

The challenge is that pharma companies usually only discover that a drug candidate either is not safe or will not work after years of effort and hundreds of millions of dollars of investment. Companies can invest $5 billion and 12-years’ time, on average, Tidona said, to create the next  “blockbuster drug.” 

“When one thinks about these numbers, he can imagine why some of these drugs are so expensive. Could an AI platform predict if a drug is really ready for the clinic and, if not, tell us what was missed before we go to trial?” Tidona asked. “This is the question.

“If drugs can be made more efficiently, medicines can become cheaper and more people in all parts of the world could afford them,” he said. 

Solving therapeutic challenges

Kahina Lang, head of Strategic Innovation at Merck Group, addresses the candidates at the AION Labs boot camp.

Based in Rehovot, not far from the the famed Weizmann Institute, AION brings together some of the biggest names in pharma, including AstraZeneca, Merck, Pfizer, and Teva, with young biotech inventors and entrepreneurs, to crack biomedical research challenges together and in a novel way. 

“We work with each of our pharma partners separately and then together to identify the top research challenges that solving would be majorly impactful for the industry at large,” AION CEO Mati Gill told Health Policy Watch. Then, AION looks for innovators and scientists to develop solutions for them.

For last week’s boot camp, 15 applicants from around the world were selected to come to Israel and participate. 

“The winner receives a $2 million investment, mentorship and access to a wealth of data for model training from these four big pharma companies,” Tidona said. “It is more or less paradise for anyone who wants to start a company.”

The concept is based on a model Tidona first developed on the campus of the University of Heidelberg in Germany, with a world-wide network of partner locations. 

If the company succeeds, the technology belongs to it but will be made available for purchase by the pharma partners. 

‘Bringing the promise of AI to fruition’

Guy Spigelman of AWS addresses candidates at the AION Labs bootcamp.

Merck’s Yerushalmi said that there have been “all kinds of efforts to make the drug development model more efficient – to expedite it,” but until now it has not worked. When computational approaches to research were first introduced, “it brought a lot of promise, but until now it has not brought as much fruit as we hoped. Now, with AI tools and the amount of data pharma and other companies can generate, we do hope we may be able to bring this promise to fruition.”

AI is one of the most sophisticated computational tools. AI analysis has brought “ingenious” solutions to other fields, like the automotive industry, she pointed out. 

“AI tools and big data have proven themselves in so many other cases, I think it will probably work for our industry too,” Yerushalmi said. “Want to make [drug development] more affordable and shorter, so we can bring drugs to the market faster and cheaper for the benefit of humanity.”

Finding the most effective antibodies 

AION will be looking for the next therapeutic antibodies.

This most recent boot camp was AION’s second one so far. Earlier this year, it invited computational biologists and biomedical scientists to propose ideas for discovering therapeutic antibodies.

“Advances in protein structure prediction, artificial intelligence algorithms, and increased availability of experimentally determined antigen-antibody structures present a unique opportunity for AI-driven antibody discovery,” AION said in a release. 

In the coming months, it will hold a third boot camp at which computational biologists, bioinformatics and cheminformatics scientists and AI researchers will propose ideas for the development of a next-generation computational platform to optimize antibodies for targeted therapies with enhanced properties, including developability or manufacturability, stability, aggregation, immunogenicity, pharmacokinetics and tissue distribution. 

“The ultimate solution is an AI platform that receives sequences of binders and generates novel variants with optimized IgG sequences, biophysical and targeting properties,” a background briefing explained. “The goal of the AI algorithm is to make an existing antibody a better drug while reducing design iterations, optimization of cycle times and lowering attrition rates.”

CEPI launches AI-based quest for beta-coronavirus vaccine candidate 

The AION initiative, while pioneering in the use of AI for drug discovery, is not the only one. 

On Thursday, the Oslo-based Coalition for Epidemic Preparedness (CEPI) announced that it would provide seed funding of up to US $4.8 million to a consortium to support the AI-based development of betacoronavirus vaccine candidates – the new holy grail for coronavirus vaccines. 

The research consortium, led by a Norway-based subsidiary of the Japanese NEC Group, which specializes in AI technologies, also includes the European Vaccine Initiative (EVI) and Oslo University Hospital. 

It aims to establish preclinical proof of concept for an mRNA-based vaccine that protects against a broad range of beta coronaviruses – rather than SARS-CoV2 alone, said CEPI in a press release.

NEC will apply its experience in the AI design of immunogens to identify novel vaccine antigens with broad reactivity against beta-coronaviruses. The lead antigens will be selected iteratively and validated in preclinical studies against known beta coronaviruses that already pose a significant epidemic or pandemic risk, such as SARS-CoV, SARS-CoV-2 and MERS-CoV.

 If the approach is successful, it may also be applicable for developing vaccines against other pathogens in the CEPI portfolio, including ‘Disease X’ – unknown pathogens with pandemic potential that have yet to emerge.

What the future holds

What the future of AI holds.

Tidona, for his part, hopes that AION will spin off 20 new start-ups in the next four to six years. 

“We are building an innovation ecosystem,” he said, noting that it is attractive to host countries as it provides a draw for youthful talent, which in turn spurs economic development. 

He added that if the model works well, BioMed X plans to export it to other countries as well.

“The focus now is on Israel as our first partner outside Germany,” Tidona said. “But in the future, we hope to have sites” in other places, too.

Image Credits:, AION Labs, Elad Malka, Twitter: @WHO,

Combat the infodemic in health information and support health policy reporting from the global South. Our growing network of journalists in Africa, Asia, Geneva and New York connect the dots between regional realities and the big global debates, with evidence-based, open access news and analysis. To make a personal or organisational contribution click here on PayPal.