Enakshi Singh is Senior Product Specialist, SAP Health

Enakshi Singh

	The enormous cost associated with drug development is one of the biggest challenges for pharmaceutical manufacturers. With an estimated cost of $2.6 billion and more than 10 years (1) to develop and test a drug candidate, the pharmaceutical industry is constantly looking for new efficiencies that will save time, money, research, and resources. One new ally that is showing promise is artificial intelligence (AI) and machine learning.

	AI is proving to have enormous potential in many areas of healthcare, including chemical research and discovery. Using large pools of aggregated data, AI can discover and learn, hence, turning data into usable and actionable knowledge. In the not-too-distant future, AI platforms coupled with in-memory computing technology will be able to provide accelerated drug discovery and delivery and also help scientists find new uses for drugs.

	The key enabler to the success of AI in drug development is sharable data. Pharma and medical researchers have vast pools of data at their disposal in the public health ecosystem. Sources of patient data include clinical trials, electronic health records (EHRs), high-resolution medical images, and genomic profiles, all of which can be useful for drug development.

	While the use of AI in clinical situations is still limited, the most practical applications for AI reside in taking on what can be considered the more mundane, data-driven tasks. Pharmaceutical researchers also have massive libraries of compounds and results from drug development testing. By using AI tools that can rapidly evaluate the extensive collections of synthetic and theoretical drug banks, the research teams-with an AI assist-can uncover hidden gems.

	Using algorithms, AI systems can discover patterns in these data pools. The AI system rapidly learns how to hone in on key information and develop hypotheses. It can refine its answers over time.  With machine learning, including deep learning, the computer is trained to figure out a problem, rather than having the answers programmed into it.  

	It is no secret that the average biomedical researcher processes huge amounts of new information. Just keeping up with all the published data on a topic is a challenge. The bioscience industry is bombarded with tens of thousands of new publications uploaded on a daily basis from biomedical databases and journals around the world. Researchers, even those with large teams, cannot begin to process all the scientific data related to their area of investigation.

	AI gives scientists and researchers the ability to correlate, assimilate, and connect existing data more rapidly. AI and in-memory computing can review billions of reported scientific research papers and abstracts and link direct relationships between the data, compiling a list of known facts. With these facts, AI can make connections that help generate a large number of possible hypotheses, using criteria set by the research team.

	An additional benefit of using AI is breadth. Researchers in the early testing phase can tap into a broader chemical space, using AI to eliminate variables faster and help researchers pick the best compounds for efficacy and safety.

	Today, the approach used by most pharmaceutical companies involves screening large numbers of molecules for promising drug candidates, and then carrying out a number of tests in the hope of finding a winner. This is an expensive exercise that takes an enormous amount of time. But what if AI could do the tedious molecular screening?

	Biotech company Insilico Medicine, for example, is working on “imagining” new cancer-like molecules with specified properties. The company is using a relatively new deep learning technique known as the generative adversarial network (GAN).

	GANs can create entirely new data that are indistinguishable from real data, by using two competing neural network models. A generative model tries to generate output that “looks like” real data. A discriminative model takes input from both the generative model and real data and tries to distinguish between them. Generative models are commonly used to create images, speech or text, but this is the first time a GAN has been used for cancer drug discovery. Researchers in Insilico’s Pharma AI division have reported that its network has used historical biological and chemical data to imagine 69 new molecules with the potential to fight cancer.

	Insilico has also used AI to help predict the therapeutic use of drugs. Its AI platform was fed experimentation data on 678 drugs and the effects they had on gene expression in three types of human cells. The AI platform developed an ability to classify drugs into therapeutic use categories, achieving 54.6% accuracy in identifying one out of 12 of the drug’s therapeutic applications.

	This type of success with AI represents a huge step for researchers, who otherwise would have to make these predictions through countless hours of experimentation. And as a bonus, the AI platform’s “wrong” answers were helpful, pointing at secondary uses for drugs that researchers had not considered.

	Less than 10% of potential medicines make it to market, meaning millions of testing hours and dollars are spent on each fallen angel drug. Eager to find ways to use that research data, manufacturers are turning to AI to find new ways to repurpose drugs that have already gone through the early stages of testing. By doing that, the turnaround to market is faster and less expensive than starting from scratch.

	AI can also build on structural and pharmacological similarities in drugs. By comparing testing results and rapidly evaluating huge amounts of data, an AI platform can find interconnections for something that hasn’t been recorded for a particular drug, eliminating hundreds of manual lab testing hours.

	Though real results are just starting to come in, AI is showing positive traction in drug development. AI platforms are helping pharma researchers identify new drug targets. AI is also showing the potential to decrease the time required to screen molecules, as AI systems can identify patterns far more rapidly than humans. It can also determine which drugs are most effective in the treatment of specific diseases. Machine learning can help pharma recover the value of repurposed drugs by finding different therapeutic applications.

	With the capability to look at 14 trillion data points in a single tissue sample, AI can help eliminate the years lost in trial-and-error drug development (2). By tapping into untapped data pools, AI has the potential to shave millions-maybe trillions-of dollars off the cost of drug development.

		R. Mullin, “Cost to Develop New Pharmaceutical Drug Now Exceeds $2.5B,”

		I. Douglas, “Cancer Drug Development Time Halved Thanks To Artificial Intelligence,” 

Enakshi Singh is senior product specialist, SAP Health.

Articles

By identifying patterns, artificial intelligence and deep learning platforms help researchers discover new drugs faster.