Tushar Satav was about to complement his career in pharmaceutical research with an MBA in data science when he was diagnosed with a massive brain tumor. Drawing on his experience, he began exploring every available treatment option for a relatively occasional type of brain tumor, vestibular schwannoma. Unfortunately, due to the rarity of this type of cancer, there was no treatment option.
Pharmaceutical companies are not incentivized to develop modern drugs when the potential patient pool is very compact. It can take many years to discover, develop and study a drug. This is followed by prolonged clinical development and a phase in which the drug must be tested for pharmaceutical side effects in humans and reviewed by government agencies. Due to these limitations, drug development requires enormous amounts of capital, and even a promising research path may fail to achieve the goal of commercialization.
Fortunately, Satava’s tumor was benign and was successfully removed. However, this experience made him think about how the conventional drug discovery process could be improved. He came up with the idea of using technology to identify existing or failed drugs for modern indications. Satava’s need for technology and business development expertise led to a meeting with co-founders Tjerk Geersing and Dr. Roland Meisel. Geersing is a technology development veteran and Meisel is a pharmaceutical business development expert with over 20 years of experience in business development and inventive dealmaking. The trio co-founded Keystonemab, a startup that uses artificial intelligence to find hidden connections between information in millions of scientific articles that can be used to develop modern treatments. Typically, drug discovery researchers can take decades to review the relevant scientific literature in a given field. However, Keystonemab software can quickly extract actionable information in real time thanks to IBM Watson’s advanced natural language processing (NLP) capabilities.
Cheaper, safer medicines – faster
In addition to discovering modern drugs, researchers can employ technology to find two or more existing drugs with synergistic effects, allowing pharmaceutical companies to experiment with modern drug combinations. This approach is called “drug repurposing” or “drug repositioning.” It provides significant improvements in treatment effectiveness, cost savings, as well as faster time to market, which can save lives in scenarios such as the current COVID-19 pandemic. It also enables pharmaceutical companies to treat occasional diseases more safely and cheaper. Drug combinations may also be less toxic and more effective than a enormous dose of a single drug.
According to Satava, the development timeline for repositioning drugs can be 30-60% shorter than for combination compounds, and overall development costs can be reduced by up to 60% because existing drugs have already undergone high-priced safety profile studies.
Satav and the team believe there are many effective treatments for occasional diseases just waiting to be discovered. However, the only way to find them is to create links between different drugs that are not usually considered complementary. Their business model is based on massive pharmaceutical research data sets recently made available to the public, which include information on drugs, diseases, drug target proteins, biomarkers and signaling pathways. Researchers can employ AI to discover meaningful semantic connections between two or more drugs with complementary characteristics, such as finding many matching needles in a very enormous haystack.
Trusted, unthreatening and accessible artificial intelligence
When Satav started looking for a technology platform, he discussed his employ case with other entrepreneurs and professionals in his industry. “They actually led me to Watson,” he says. “Watson is a proven technology. Many pharma startups use Watson, so I don’t have to explain to customers why we use this particular technology because everything is really proven.”
Another factor was Watson’s safety. Pharmaceutical customers require a high level of security for their data due to the highly regulated nature of the industry. “We work with pharmaceutical clients and data security is very important to us,” says Satav.
Third, Satav chose Watson because he is not an AI expert by training. “It’s very user-friendly,” he says. “I’m not an artificial intelligence expert, I’m more of a wet lab scientist. I was used to working in drug discovery labs, extending to big data analysis, but I was not an artificial intelligence or software specialist. So user-friendliness was important to me.”
Keystonemab uses Watson Knowledge Studio and Watson Natural Language Understanding (NLU) to build and train custom models that identify millions of entities and relationships, determine the strength of those connections, and ignore the weakest links. It took Keystonemab about a year to build the solution, and the company plans to spend another six months bringing it to market.
Disrupting Massive Pharma and more
Keystonemab is marketed primarily to chief scientific officers, chief development officers and other senior executives at compact and enormous drug discovery companies, but the goal is to provide a platform to commercial drug manufacturers, laboratory scientists, physicians and other users further down the supply chain. For now, the team is focused on presenting discoveries to partners that they wouldn’t be able to discover with their current in-house research staff.
But first, Satav and his co-founders must convince decision-makers that artificial intelligence is not just noise. “The pharmaceutical industry is not very aware of artificial intelligence, and this is because it is conservative in using new technologies in which it is not an expert,” says Satav. “So you have to explain it to them or it has to be simple enough for them to understand that it’s not a black box. It’s simply technology that can significantly improve their everyday working lives.”
Satav and Meisel’s pharmaceutical expertise goes a long way in convincing these potential customers.
“If you don’t know their language, it’s very difficult to convince them,” he says. “Historically, my industry is not used to change, so it can’t be easily disrupted, but if you’re part of the same community and can explain the technology to them in their own terminology, they’re more likely to adopt it.”
Keystonemab’s long-term goals are to turn its insights into modern treatments and ultimately expand applications beyond pharmaceuticals into the nutrition, healthcare and other industrial sectors. The same ability to find connections between drugs can theoretically be used to draw conclusions from the scientific literature on the production of all kinds of sophisticated chemical products, as well as how nutraceuticals can support patients, especially in the case of nutritional deficiencies caused by, for example, stomach diseases or diabetes. A similar approach is also envisaged in the bioprocessing industry, as the dependence of cells on media for culture is comparable to that of patients suffering from a lack of certain trace elements or other nutritional deficiencies.
But for now, Keystonemab is solely focused on drug discovery and helping reduce the risks associated with the highly risky process drug research companies undertake when developing modern treatments. This effort is a fantastic employ case for AI search, demonstrating how finding ways to search documents quickly and more accurately can literally save lives.