Achieving AI: Pharma’s Digital Transformation Pathway

By Allison Proffitt

June 10, 2021 | What does it mean for a pharmaceutical company to be digital and how do we get there? That’s what Reza Olfati-Saber, PhD, Global Head AI & Deep Analytics, Digital & Data Science R&D, at Sanofi tackled yesterday at the DECODE: AI for Pharmaceuticals forum.

A digital pharma company is agile, Olfati-Saber argued, enabling it to discovery drugs faster and develop and manufacture drugs more efficiently. Olfati-Saber pointed out that of the four first-movers in COVID-19 vaccine race—BioNTech/Pfizer, Moderna, AstraZeneca, and Janssen Pharmaceuticals—all have reputations as digitally-advanced companies.

“The one thing all four companies have in common is all of them digitally-advanced biopharma companies. The last two, among the larger pharma companies, happen to have very advanced AI and ML capabilities,” Olfati-Saber said.

“Digital” can be hard to pin down, Olfati-Saber conceded, and he observed that many groups are eager to jump in and claim AI expertise. Lawyers seek to define “ethical AI” but don’t generally take medical ethics into account, he said, while management professors claim to roadmap the “digital transformation journey” without any industry-specific insight. Even “digital” is defined in the most convenient way for each industry.

Olfati-Saber narrowed the scope to discuss the meaning and architecture of digital transformation specifically for pharma R&D.

Digital Architecture Models

For pharma R&D, the digital transformation narrative can be illustrated as a pyramid architecture, Olfati-Saber said. The traditional pyramid has computing (cloud, infrastructure) as its wide base, advancing through applications (data storage, app development, security), data (data governance and security), AI policy (quality and ethics), analytics (data analytics and visualization), and machine learning.

Olfati-Saber views the lowest three layers of this pyramid—computing, applications, and data—as the foundational digital layers. The first two are technical requirements. Together, along with the data layer, these three confer AI enablement. These competencies must be in place for a company to be AI-ready. AI policy, analytics, and machine learning make up the true AI capabilities for an enterprise, and these sit at the top of the pyramid.

But contrary to narrow expertise of “AI experts” from law firms and management schools, Olfati-Saber argues that true digital transformation of a pharma company requires expertise from four quadrants. Both technology and management expertise are required to build the solid digital enablement foundations, and scientific and legal expertise combine to drive AI.

In fact, Olfati-Saber argues that it is “practically impossible” to expect a Chief Data Officer to know the entire quadrant well enough to facilitate a digital transformation. Instead, he argues for both a top digital expert and a top AI expert working together. “Anything else wouldn’t do the job,” he said.

Development Pathways

It’s a complex schema, and Olfati-Saber proposes a four-phase pathway for development. Start by establishing the technical foundations, then add the needed data, the AI tools, and finally fine-tune the enterprise AI policy.

It’s a slight rearrangement of the traditional pyramid view, moving AI policy to the final phase of development or pinnacle of the pyramid and grouping analytics and machine learning together below.

The rearrangement reflects what Olfati-Saber sees as the hardest part of the digital transformation—the biggest stumbling block for companies.

“Despite the fact that many large tech companies have gone through the first two phases of transformation really successfully, they’re struggling to go through the last phases of transformation,” he said. “Part of the reason is that there seems to be some sort of conflict between the business models of some tech companies and some of these AI and data privacy-related policies.”

He alluded to Google’s recent dismissal of company ethicists when their ethics findings—presumably in a paper submitted to an industry conference—didn’t align with the company’s goals.

“It’s not easy to simply put together a committee and expect them to form the quality assurance and ethics principles of AI,” Olfati-Saber said. “This is a very challenging task, just as challenging as any of those other three.”

Deep Digital Transformation

When a company has achieved all four phases, Olfati-Saber said, they’ve undergone what he calls a “deep digital transformation.” And it’s a worthwhile process, he argued. He outlined many examples of where AI can impact the pharmaceutical business: digital pathology, AI-based drug design, multi-omics analysis, digital health, digital manufacturing, AI-based regulatory approvals, and more.

In his own estimate, as an example, Olfati-Saber argued that the AI-enabled cost savings per image suggests digital pathology is 2,500x cheaper than standard pathology and 60x faster, even when the digital tools are simply aiding pathologists, not replacing them.

“The real reason pharma companies or investors out there are interested in applying AI and investing in AI for pharma is not because it’s a fancy tool, or because it’s fashionable. It’s mostly because it generates massive returns in agility, scalability, and cost savings,” Olfati-Saber said. “These are the true reasons why a pharma company would want to become AI ready, go through a digital transformation, and have AI capabilities.”