Winter is coming to the artificial intelligence (AI) industry. While it's springtime for investment and technological advancement, gray skies hover over the talent pool.
There are now about three times more firms using AI in their drug discovery process than there were eight years ago. A survey by Deep Knowledge Analytics reveals 44 drug discovery companies using AI and machine learning in 2011. By the end of 2018, there were 125 such companies. Investment in AI for drug discovery steadily and impressively increases. Deep Knowledge found $76 million invested in AI for drug discovery companies in 2014 and by the end of last year, it soared to $1.1 billion.
So, capital investment? Check. Tech advances? Check. Qualified personnel? Hmm. More AI-savvy staff is needed to fill the jobs all this progress created. And not just bodies but employees who understand the new terrain. Companies know it, and they are hiring.
Sometimes the skills gap is a factor for not progressing. Tufts University's Center for the Study of Drug Development found that safety and environmental concerns are a huge reason many companies don't use AI technology. Budget constraints and mistrust of unvalidated technologies also scares some off. But the biggest factor holding them is lack of skilled staff, as 58% of the companies surveyed by Tufts noted.
The human hands on the AI buttons are there, but it can be a fraught process finding the right person with the right qualifications. Tufts found that 59% of respondents plan to expand their AI staffs throughout 2020, with the largest staffing increases in the data scientist, computer scientist, IT specialist and AI architect slots. Some of those new positions are barely understood by companies who need the talent.
"Machine learning, data science, didn't exist until five years ago," Tina Larson, chief operating officer at Recursion Pharmaceuticals Inc., told BioWorld MedTech. "It's a very new field, and there are only a few people experienced in this field."
Secret sauce: multidisciplinary-ism
For Recursion, the secret to hiring the right people for the right AI job is "multidisciplinary-ism," according to Larson, who admits that may not be a real word but it does get the point across. New hires need to be able to think and work across disciplines for the process to work well. Recursion hires people straight out of their doctoral or post-doctoral work. Job titles include chemists, research biologists and data scientists. More than half of Recursion's workforce has a doctorate, master's or another graduate level technical degree. Forty percent are high-science professionals (biology, chemistry), 35% are high-tech professionals (data science, software engineering) and the remaining 25% are business professionals (product management, intellectual property, etc.).
Larson said many prospects may have experience at other biotech and biopharma companies, but few of them understand machine learning in particular. The secret is getting them out of their niches and into their colleagues' niche so they can trust each other.
"For computer scientists, we train them to understand enough biology so they are computationally savvy, and we train biologists to understand how to be computationally savvy," Larson said. "Increasingly, chemistry is being computationally driven."
Academic institutions are fertile hiring grounds, particularly from the math, chemistry and biology departments, which, Larson finds, don't always interact.
"What's surprising about working in a university system is that the whole concept is not very multidisciplinary," Larson said. "The democratization of information has changed science. We need scientists and engineers [who] have learned how to learn and have a good foundation across multiple disciplines."
In what can be a virtual business, much of the success of AI and machine learning drug development hiring is often driven by geography. Salt Lake City-based Recursion finds 37% of its workforce is from out of state.
Abbvie Inc. uses a global team in its Development Design Center that uses predictive analytics and real-world data to drive innovation in clinical trial design. It looked for job candidates with backgrounds in computer science and statistics. They needed to have the ability to evaluate large data sets, but the company still values traditional scientific and medical backgrounds.
"The skill sets of some of our teams have changed in the past few years," Abbvie's Kyle Holen, who heads the Development Design Center, told BioWorld MedTech. "Our team has an impressive track record of translating science into effective medicines that have advanced the treatment of diseases, and as the clinical development process continues to evolve to integrate digital health and become more data driven, people with backgrounds in programming, machine learning and artificial intelligence have been critical to our success."
Again, collaboration at the heart of the process, Holen said, but it's not everything. Regardless of the specific area of expertise, having that core knowledge of the process of the discovery and development of medicines helps the team members work together toward the same goal.
"In our industry, more important than cross training in these areas is knowledge of the clinical research process and how an investigational medicine gets from the lab to a patient," Holen added.
Mind Foundry Life Sciences advisor David Bennett cited research from Deloitte LLP showing productivity and R&D returns in biopharma companies have dropped to their lowest levels in nine years, that companies constantly question whether staff should be in-house, outsourced to a smaller company or involving academia.
"With talented data scientists in scarce supply, the skills gap is continuing to pose challenges to life sciences organizations," Bennett said. "Existing data science departments do not have a wealth of data scientists, so their talents – and workloads – are reserved solely for the most business-critical and time-sensitive tasks, particularly in the R&D space."
Deloitte also found that 95% of companies surveyed plan to use machine learning to support real-world data analyses in the next few years.
Beyond all the promises that have been made for AI in drug discovery, Bennett continued, the real transformation in productivity in life science companies value chain will be wrought by augmenting the existing workforce with AI and moving beyond the realm of the specialist data scientist. Machine learning can be harnessed to find and enroll patients in the most suitable trials and facilitate the entire patient journey.
Abbvie and Recursion aren't alone in beefing up their AI staffing. The Tufts study shows that 59% of biopharma and biotechs surveyed plan to increase their staff sizes in the next year or two, that 19% plan increase in the next three to five years, and 7% plan an increase in more than five years. Only 15% of companies surveyed plan no increase in their AI rosters. Small increases were anticipated by 61% of the companies surveyed, moderate increases by 28% and large increases by 11%.
Top of the list: data scientists
By far, the Tufts study found, the role seeing the largest increase, 82%, is data scientists. Computer scientists come in at 59%, closely followed by IT specialists. AI architects will increase by 50% with statisticians right behind them.
Academia has traditionally been home to most of the top research minds in pharmaceutical and healthcare AI, according to Dmitry Kaminskiy of Deep Knowledge Ventures, but they could also be AI companies or pharma corporations, with tech corporations coming in last.
Brooke Clark, Recursion's senior director of talent acquisition, finds talent from all over the world. She runs the in-house talent acquisition team while using recruiters to fill senior and executive level positions.
"Really smart people come from everywhere," she told BioWorld MedTech. "The question is how can we take advantage of a global talent pool."
Clark helped build the company with an equal emphasis on tech and life science using individuals with AI expertise. She interviews and evaluates candidates with strong emphasis on an interest in collaborating. Curiosity about other disciplines is a must.
"I found that we have to build relationships and establish trust between different groups right away. They're subject matter experts in their own fields, but they must relate to each other," she said. "We're looking for individuals who are not only strong in technical expertise but who have a strong desire to collaborate with those who are widely different [from] what they're used to collaborating with."
While the demographics of AI staff tend to be fairly young, Clark hired a blend of professionals from all phases in their careers. Some are just out of post-doctoral programs and are there with those who have spent years in computer science, biology and the pharmaceutical industry.
"That's the beauty of these teams," she said. "We're doing a thing that's never been done."