Despite major advances, drug development remains a costly and time-consuming endeavor. One 2020 study pegged the average cost of bringing a new drug to market at $1.3 billion. And the process of discovering and preparing the drug for sale can take from 4 to 10 years.
The rise of artificial intelligence (AI) may help streamline drug discovery and reduce the time and money researchers will need to develop new drugs.
The pattern-finding and predictive abilities of AI make it an excellent fit for businesses wanting to develop more effective predictive models and automate tasks that have proven difficult to automate in the past — like early laboratory testing of new drugs.
In drug development, AI adoption has had a noticeable impact on every step of the drug manufacturing process, from discovery to testing to mass manufacturing. A few key AI innovations have been the primary drivers behind the adoption of AI in the industry.
Leveraging AI to speed drug discovery
In drug discovery, a significant portion of the money allocated is spent investigating dead ends — molecules that, ultimately, prove ineffective or have side effects that make them impractical for widespread use. Tools that can streamline this process, reducing the risk of dead ends, could significantly reduce the time and financial costs of developing new drugs.
The pattern-finding abilities of AI make it an excellent tool for sifting through potential drug candidates to identify compounds or existing treatments that may produce the response a research team is looking for.
The process of using AI in drug discovery typically begins with a research team generating millions of novel compounds and molecules, similar to those already in use for a particular purpose, like cancer treatment drugs.
The team then uses an AI algorithm that can predict the properties and medical effects of those procedurally generated compounds, revealing effective drugs that may never have been uncovered with a more conventional approach.
The top candidates identified by the AI system can then be subjected to the typical batteries of tests for new drugs.
Early applications of AI in drug discovery
For example, this approach has already been used to develop what is believed to be the first drug created entirely by AI. It’s a new adjuvanted flu vaccine, currently under trial in the U.S.
A research team from the Australian Flinders University, headed by Director of Endocrinology Nikolai Petrovsky, used an AI tool known as the Search Algorithm for Ligands (SAM) to discover a new compound that could work as an adjuvant for flu vaccines.
Adjuvants, substances that increase or modify the body’s immune response to a vaccine, are increasingly common in flu vaccines and are sometimes recommended for immunosuppressed or elderly patients.
The research team first developed a program called the “synthetic chemist” to generate “trillions of different chemical compounds.” These compounds were then fed through the SAM algorithm, which selected compounds whose structure suggested properties that may help the human immune system fight influenza.
Drugs selected by SAM were then tested on human blood cells. When the immune response provoked by these new adjuvants suggested they may be even better than existing drugs, the team moved on to animal trials. In 2019, following the success of those trials, the team began preparations for Phase II human trials, sponsored by the National Institute of Allergy and Infectious Diseases.
Similar technology has already been used for different drugs. For example, AI was used by the German biotechnology company Evotec to automate a full-stack approach to the drug development of a new anticancer molecule. The company’s proprietary “Centaur Chemist” AI design platform is able to sift through millions of potential molecules, then select 10 to 20 to synthesize and test in a laboratory.
Potential impact of AI in drug laboratory testing
AI algorithms could also help streamline early testing of compounds in addition to determining good potential candidates. Lab automation has proven to be particularly difficult, but AI is often an effective tool for automatically managing tasks that conventional approaches have struggled to automate in the past.
Machines like flow chemistry reactors, for example, are essential for a number of processes necessary to create and test new drugs, like chemical synthesis and impurity profiling. AI-powered lab automation platforms can help researchers automate these tasks, increasing the number of novel compounds they can synthesize, purify, and test in a given timeframe.
Techniques like high-throughput screening use automated laboratory equipment to evaluate large numbers of biological compounds against certain biological targets — like an immune response.
These automated tests can enable researchers to experiment with hundreds of thousands of agents in a way that is both reliable and reproducible.
Streamlined testing practices like these can make testing novel compounds identified by AI much more practical — even when the AI only narrows down the testing pool to a few hundred potential drugs, rather than a few dozen.
Repurposing existing therapies with AI
The same analytical approach researchers are using to discover new compounds can also be effective in identifying alternative uses for already approved therapies.
During the COVID-19 pandemic, for example, the London-based firm BenevolentAI used AI to identify the rheumatology drug baricitinib as a potential COVID-19 treatment. With the use of an algorithm, the company was able to move from an idea to “actionable research” in just 48 hours.
This kind of research could help companies fast-track the study of existing drugs to find potential alternative applications. These therapies allow researchers to begin with an understanding of how the potential treatment could work and what kinds of side effects patients may experience.
How AI may impact drug discovery
Artificial intelligence has the potential to transform the often time-consuming and expensive process of drug discovery.
New algorithms can be used to both generate and sift through millions of potential compounds. Other algorithms may help improve lab automation, allowing research teams to automate the processes of discovering these compounds and testing them in the lab.
While the use of AI isn’t widespread yet, it’s already been used to support the development of several new drugs. As these drugs move through clinical trials over the next few years, researchers and businesses may be inspired to adopt AI technology for drug discovery.