The adoption of artificial intelligence (AI) has become a strategic priority among pharmaceutical companies, who are increasingly using it and machine learning (ML) predictive models to accelerate drug discovery.
At Pharma IQ’s second edition of Transforming Drug Discovery Through Artificial Intelligence, speakers from companies including AstraZeneca, Bayer, Recursion Pharmaceuticals and Sanofi, outlined how they are using AI to drive efficiencies and cost-effectiveness.
Bayer: Fine tuning research compounds with AI and machine learning
Dr Alexander Hillisch, head of computational molecular design at Bayer, spoke about digitalization in the pharmaceutical industry, the digital technologies deployed at Bayer and the impact of competition molecular design on the development pipeline.
He noted that while digitalization is truly disruptive for many industries, it is lagging behind in pharma because of the complexity of biological systems in the early phases of drug discovery.
Digitalization is interesting, he said, because “calculations are quick, inexpensive and offer a chance to get better quality results and do things that have not been possible so far”.
“We are in a transition in pharmaceutical research, from purely trial and error on one hand, to purely engineering on the other side, and we are somewhere in the middle today,” he noted.
Hillisch took the audience through the intricacies of using AI and machine learning phase property prediction to fine tune the compounds they are researching. He added that digitalization is rewarding but requires significant investments.
“Pharmaceutical companies hold all the data and excellent workers in digital sciences but they cannot do it alone and need partners in the digital arena and in academia,” he stated.
Recursion Pharma: Discovering links between gene mutations through AI
Christos Nicolaou, Director in Digital Chemistry at Recursion Pharma, explained how his organization is using digital chemistry tools to optimize drug discovery. In particular, Recursion has been building phenomaps and using these to identify links between all the potential mutations in genes – something that would be impossible without AI.
“We have the goal of shrinking the time each discovery cycle takes, but also cutting the number of cycles down,” he told us in an interview ahead of his session. “It is not just making things faster, it is also designing intelligently using all the AI and ML models that we can build –rather than make 10, 20 or 30 cycles we can get to the right kind of compounds in five.”
AstraZeneca: Extracting value through data mining
On day two, Shameer Khader, Senior Director in AI/ML, Data Science, Digital Health and Bioinformatics at AstraZeneca, opened with a session on how to extract value when mining clinical trial data.
Covering the various challenges scientists face when analyzing data from large-scale studies, Khader said that a lot of data stays locked within the databases and never gets used.
Khader and his team wanted to understand the reasons for patients dropping out from trials. They found that this information was not readily available, which led to a project called Aggregate Analysis of Clinical Trials (AACT).
Using the ClinicalTrials.gov database of clinical studies conducted around the world, Khader explained that they deployed a classic text mining approach followed by vector space modelling.
“It is an algorithm that can mine across the data using semantic reasoning, so out of 10,000 reasons we are able to bring it down to 1000 and a more comprehensive list of terms,” he said. “We believe there's a lot of data out there, but it is like gold mining.”
Sanofi: Developing new pharma molecules with quantum chemical methods
In the final session Jan Wenzel, a principal scientist in Preclinical Safety and Digital Toxicology at Sanofi, presented a case study that explored how predictive computational methods are enabling the development of new pharmaceutical molecules.
Wenzel explained how Sanofi is using quantum chemical methods to predict phototoxicity, which occurs when certain drug molecules interact with UV light and can lead to sunburn and skin eruption.
“The predictive quality of this model is very high and it really helps us to identify the phototoxicity of a compound in the early stages,” he told Pharma IQ ahead of his talk. “Then we can easily react and provide guidance on which part of the molecule needs to be transformed to reduce the risk of phototoxicity.
Quick links:
- Three ways AI is speeding up drug discovery
- How Sanofi is using AI to accelerate time to market for new drugs
- What is the digital lab?
Get exclusive access to member-only articles, reports, videos, interviews, webinars and other premium content from industry experts and thought leaders by signing up to Pharma IQ here.