Today, most new drugs are often developed through drug design using combinatorial chemistry. Artificial intelligence provides new insights by making it possible to determine which substances are most interesting to test pharmacologically.
In the context of the search for drug candidates, there are several stages of development, particularly in the choice of active ingredients. Combinatorial chemistry is most often used. Given the volume of data to be processed, artificial intelligence appears to be a solution for optimising research.
Today, a company like Iktos offers concrete solutions to support R&D teams. It has forged partnerships with many pharmaceutical companies to help them accelerate and optimize drug design phases (Janssen, Merck, Servier, etc.).
Another example is the partnership between the University of Toronto and the start-up Insilico Medicine, which used artificial intelligence to identify a potential treatment for fibrosis in forty-six days.
In concrete terms, the artificial intelligence solutions deployed generated 30,000 models of molecules in just 21 days. After a selection of 6 of these molecules synthesized in the laboratory, 2 were tested on stem cells to finally test the most promising molecule in animals after only 46 days.
Historically, the research and development of new medicines has been based mainly on laboratory (“in vitro”) experiments and then on animals and humans (“in vivo”) via clinical trials.
Digital technology and artificial intelligence are opening up a new voice with digital simulation: “in silico”. We are talking about the third pillar of drug development. This approach makes it possible to characterise and predict the toxicity of a drug candidate, or even its efficacy, even before it is tested in “in vivo” clinical trials. We can simulate the effect of a molecule on a pathology and on a patient population. This makes it possible to refine the clinical development strategy and to help reduce the time and costs involved in the development of molecules.
A concrete example with the start-up Novadiscovery, which specializes in quantitative modeling and simulation applied to R&D.
Drug optimization using AI
Artificial intelligence optimizes the upstream phases of research but also allows the development of new molecules in a shorter time frame.
A first molecule developed thanks to artificial intelligence will be tested in the coming weeks on humans in Japan to treat Obsessive Compulsive Disorders (OCD). Developed by Sumitomo Dainippon Pharma in collaboration with Exscientia, this AI has made it possible to accelerate the development process from 5 years to 12 months.
This molecule, named DSP-1181, was developed using algorithms that screened compounds potentially usable to create this drug and compared them to a huge database. “If AI was used to diagnose patients and analyze data and scans, this is the first direct use of AI in the creation of a new drug,” said Professor Andrew Hopkins, CEO of Exscientia in a release.
In another recent innovation, researchers at the Massachusetts Institute of Technology (MIT) have succeeded in using a machine learning model to identify a molecule capable of countering one of the most formidable pathogens: the gastrointestinal bacterium Clostridium difficile (C. diff), which is wreaking havoc in hospitals. Developed by Regina Barzilay, a professor of computer science and artificial intelligence at MIT, this Machine Learning model is based on a deep neural network, a system trained on more than 2,500 chemical structures.
The molecule called Halicin, tested on a mouse, proved to be effective with a very different structure than existing antibiotics. It was found in a database where it was initially identified as a possible treatment for diabetes (source: The Journal cell).
A first concrete achievement of AI in the creation of drugs that will become widespread in the coming months.
The increased computational capacities provided by artificial intelligence and the acceleration of development processes are opening up a new era for research & development, particularly within pharmaceutical companies, with shorter lead times and reduced costs. R&D 4.0 is on the way!