In drug development, the process from idea to marketing authorisation takes on average 10–15 years. Early-phase research largely determines how difficult or easy further development of the drug will be.
"The use of AI is interesting because it will give us better drug candidates for further research and speed up the whole process," say Mikko Karjalainen, Head of Discovery Technologies, and Julius Sipilä, Head of Medicinal Chemistry at Orion's R&D.
Orion's R&D produces new original molecules for drug development. The main goal is, of course, drugs that are as effective as possible and safe for patients.
“AI enables us to optimise molecules better than before, and it will allow us to detect any adverse effects even earlier,” Sipilä says.
Robots tackle the literature
Orion has engaged in computer aided drug development since the late 1980s. In practice, this has meant the production of physics-based models of molecular structures and simple machine learning.
“With AI we are referring to increased computing capacity and graphics processors that have allowed us to handle even larger data masses and produce more precise models,” Sipilä explains.
Orion has been using AI for about five years. At its simplest, it is used for a basic research function: literary searches. Robot tools can instantly find connections that would take a researcher endless hours.
Machine intelligence to assist imaging
The team led by Mikko Karjalainen is responsible for early-phase biological screening in drug development, and AI is already used to a certain extent in this.
“The role of imaging is growing in drug development, including in early-phase screening for drug candidates. We teach certain characteristics of images to AI, which then goes through hundreds of thousands of digital images, which would be impossible manually,” says Karjalainen.
The data from the images is used to evaluate whether a drug has worked or whether it has potential side effects.
AI is used in designing the best possible molecules
Sipilä's group is responsible for the computational work in research projects. It uses AI to predict certain characteristics of molecular structures that affect whether a substance can be developed into a drug. Prediction has been done before, but with AI it is just faster and more precise.
Generative models are a completely new thing, in which algorithms are made to design molecules: the model is taught to make molecular structures that have certain, desired properties.
"You can leave your computer to work on this overnight, and in the morning, the best possible plans are ready for the researchers," Sipilä says.
Researcher steers and interprets, AI tackles the routines
Karjalainen points out that AI speeds up and enhances the work, but that it also eliminates human error. Machines don’t have bad days and they can work for as long as they have enough power.
"The use of AI is interesting because it will give us better drug candidates for further research and speed up the whole process," say Mikko Karjalainen, and Julius Sipilä.
“AI also reduces unnecessary work. In the future, even before any tests have been done, we may know what a drug is like: whether it is safe and whether it has any weaknesses,” he says.
But AI is not replacing humans, quite the opposite. Humans still need to tell AI what to do and interpret the results.
“AI is there to support the researcher. It is not a substitute for a human being, but instead takes care of mind-numbing and time-consuming routines,” says Sipilä.
Industry is developing at a rapid pace – cooperation is essential
Karjalainen and Sipilä believe that the AI revolution is only just beginning.
“AI is already being used for routines in individual blocks, The goal is to bring the whole cycle of development work under the scope of AI - and this is the direction we are moving in,” Sipilä says.
New developments require close cooperation with research institutes and companies. For example, Orion is looking into how to extend its use of AI with the Finnish Centre for Artificial Intelligence.
“We do not have our own competence in AI, so cooperation is essential. Over the longer term, competence will, of course, also be developed within Orion,” Sipilä says.
Orion's work will also benefit others.
“The goal of academic research institutes, in particular, is to promote the whole field. We get the tools and they get publishable results. And the research will advance,” says Karjalainen.
Strong support from Orion's management - AI opens the door to new horizons
The drug developers say that Orion's management strongly supports the use of AI.
“Boosting efficiency is one benefit. The humans can focus on the things that really matter while the rest of the work is automated,” Sipilä says. "I believe that the management also sees this as an opportunity for a revolution: we will be able to do entirely new things that we are unable to anticipate yet."
AI also offers a valuable competitive advantage over international pharmaceutical giants. There are significantly fewer resources in the area of trials, but Orion could be on the same level when it comes to AI applications.
Karjalainen points out that pharmaceutical R&D is not the only area where Orion utilises and develops AI and that its use is being advanced in many areas.
"For example, AI can allow us to monitor patient care even better," he says.
Text: Sanna Jäppinen
January 21, 2020