More AI But Fewer Process Chemists?
By Louis Garguilo, Chief Editor, Outsourced Pharma
Artificial intelligence (AI) has made its way to the mid-section of our new-drug life cycle.
Reports from around the industry indicate AI now directly enables route-scouting and data scientists, who today usher discovery programs through the rigors of development.
This also includes a technology-enhanced fighting chance for new compounds to become optimally profitable commercial products.
And this is taking place at drug sponsors and externally at your CDMOs.
How?
For example, as we document earlier, Elsevier, an information analytics business, has created Reaxys, a chemistry database with a billion-plus chemistry facts enhanced by its AI partner, Pending.AI.
What we didn’t cover directly in part one were questions such as:
- What are the potential impacts on the current workforce with the introduction of productivity-boosting AI?
- Will the future process-development workforce have different academic backgrounds and experience in different areas?
- Will AI alter the overall sponsor-CDMO relationship and decision-making process?
Dr. Jürgen Swienty-Busch, Director Product Management Chemistry, Elsevier Simon Wagschal, Associate Director, Small Molecules R&D; and Aaron Johnson, Cheminformatics Data Scientist, Small Molecules R&D, Lonza.
Changing The Guard In Development
“It is important to note we are still in the infancy stages of AI," Johnson starts us off. "Not only AI applications for our development purposes – or even throughout pharma – but in every industry today.”
Denoted by his title, Johnson is in fact a cheminformatics scientist, and he and I attempt to draw a corollary between the advent of that in silico discipline, and the introduction of AI, on chemical development and those practicing it.
I mention to Johnson we don’t hear the word “cheminformatics” as much as we once did.
“I wouldn't say it's an old term,” he says, “but we can say someone who does cheminformatics today would be better described as a data scientist fully aware of the chemical domain.”
He adds:
“I don't think there are many people who are doing cheminformatics now who don't have the skills of a data scientist, and who are not creating models and related kinds of activity.
“The definition of the term has most likely changed.”
Changed, and layered in with a good dose of pharma-industry advancement, and AI.
If Johnson was to recruit a new hire to assist him, what background would he look for?
“Broadly speaking, someone that has experience in both chemistry and data science, no matter the title,” he responds.
Swienty-Busch interjects here:
“Moreover, it looks like that augmentation of chemistry and data science with the insertion of AI methodology, is now attracting more young talent to our industry than in the past.”
“A decade or so ago here in Germany, the focus on chemoinformatics was slowing down. Now with AI in this field, I see younger people coming up, like Marwin Segler [Microsoft Research, UK,] Philippe Schwaller [EPFL Switzerland] and Connor Coley [Massachusetts Institute of Technology (MIT), U.S.], and many others who might be described as interdisciplinary.
“I believe,” finishes Swienty-Busch, “biopharmaceutical development has become quite attractive.”
AI Is Not Replacing Chemists
Wagschal, a process chemist through and through, insists AI applications to the methodology of small-molecule route scouting will not cull the ranks of professionals.
In fact, he says, their value increases because, conversely, “good process chemists are needed to enhance the technology.”
“It's more than just ‘query an AI-enabled software or database,’” he explains.
“You need skilled organic chemists to run these tools. In some cases, the software isn’t able to perform a full retrosynthesis on the whole compound. You need a skilled person to know where to cut.
“The concept is to advance our industry by combining AI and the brains of humans.
“We have to govern – and drive – the AI so it helps us collect data, review the literature, make the chemical connections faster; scientists must perform the experiments and produce the actual material.”
Adds Swienty-Busch:
“I think it was somebody from IBM who said that AI is not replacing chemists, but chemists using AI are replacing chemists who aren’t utilizing it.”
Let’s boil the above down to impacts on your outsourcing. A key question of your CDMO should now include:
Are your process development scientists AI-enabled and data-savvy, and if so, how is that applied?
Customer Expectations
Today, few believe AI will fundamentally alter the CDMOs customer involvement in the developmental direction of their programs.
Wagschal says, for example, the enhanced service at his company is “designed for small biotechs to stay engaged with us,” and adds:
“But at the same time, there is now less worry about the ability to develop a realistic and secure supply chain as we develop a program.”
“Combining AI with proprietary databases and the knowhow of the chemist, the customer is ensured nothing is being missed.
“AI-enabled scientists can put all factors on the table, and rank the candidate routes to decide what direction we want to go in first.” And, he adds, make those decisions and get going on the projects more quickly than in the past.
According to Wagschal, integrated AI allows sponsors to work closely with development scientists to consider improvements in production schemes at three entry points: preclinical, clinical, and even commercial.
With this new speed for investigating and analyzing more productive steps to a synthesis or an entire new synthetic route; and with the (AI-supported) insurance there is a viable supply chain to gather materials and reliably produce product (see part one), process development takes a giant leap forward.
That can only improve the relationship between sponsor and CDMO.
But a final word of caution:
All outsourcing relationships start and progress optimally with the CDMO’s clearly understanding the sponsors needs and concerns.
Only then can you take advantage of these advanced “retrosynthetic analysis tools,” and says Wagschal, “can the CDMO launch a call for ideas to our entire chemist community on behalf of the customer.”
And when that call goes out, indications point to a new generation of development scientists will be there to help their more tenured coworkers.