Whenever Proventa holds a Strategy Meeting around the world – regardless of meeting topic – the subject of AI comes up. It really is so important in pharma right now, with the potential to solve some of the sector’s big issues. But no-one quite knows where it’ll go next. 

We’re not content to leave these questions unanswered. We looked at a major issue in pharma today – the decline of R&D productivity – to see how AI could help. We spoke to field experts to learn about the problem, what AI is currently doing, and what will change from 2020. 

This is an abridged and shortened version of what we found. The full version of the white paper is available at the bottom of the article. 

If you’d prefer to discover these insights firsthand, join us for our first R&D Strategy Meeting this year in London on 29 and 30 June. 

How Will AI Change R&D in 2020?

Acquisitions

Big pharma companies are starting to build their own internal expertise. There have been few acquisitions of AI startups by big pharma in the past five years, however. This is despite credible demonstrations that AI can accelerate at least a small part of pharma R&D. Until these acquisitions begin, Dr. Alexander Zhavoronkov, CEO of Insilico Medicine, argues that there will be no ‘Year of AI’.

Acquisitions are the easiest means by which new technology and innovation can be brought into a company. According to Dr. Zhavoronkov, executives acquire companies with clinical-stage assets first, despite being perhaps less innovative than other AI companies.  He added that 2019 is the first year where startups are holding credible validation exercises in multiple areas. 

Take-up of AI and ML algorithms in the industry has slowly increased over the last few years. Studies still show, however, that fewer than 5% of healthcare organisations have invested in AI technologies. Despite AI’s use today, only as take-up increases and more companies invest will the technology’s potential be seen.

Shortage of Talent and the Move to Top-Down Skillbases

One of the major difficulties pharma companies face in the coming years is a dearth of industry specialists. More traditional IT and AI companies have acquired many of these. At present, only around 15.6% of AI-driven drug discovery companies’ staff are AI experts.

To combat this lack of expertise and rectify other issues within evolving pharma companies, Dr. Zhavoronkov noted the need for Chiefs of AI to take a more prominent and strategic role in a pharmaceutical company. Often, he said, when companies select their Chief of AI they look for an individual who is embracing AI for the stratification of trials or patient sub-populations, or who excels at text data analysis. What’s more crucial is an AI expert who is able to look at the situation from end to end. They must have the power internally to transform drug discovery processes to incorporate large-scale changes. “You need to put the chief of AI as CEO or CSO of the company.”

A recent study backs up this idea. Only 3% of CEOs and board members in the U.S, Germany and Japan had any experience in both AI and pharma. Their companies were, however, expected to outperform the market due to this knowledge.

Deep Learning

Dr. Peter Henstock, AI & Machine Learning Technical Leader at Pfizer, noted that predictive algorithms have existed in pharma for over 15 years without much change. Deep learning can perform such predictions more accurately – and the technology can be applied to vastly more applications. These include literature and patent mining, image processing, biology and chemistry problems.

Deep Learning consists of a number of hidden layers between input signal and result. Each layer operates independently of its peers but simultaneously. Currently, deep learning is around 10% more accurate at analysing data than the average physician. 

New areas affected include image processing, which is vastly more possible with the AI-granted ability to analyse every single cell on every single slide produced. New algorithms can show details of elements missed by scientists, identify obscure patterns, and determine how individuals are rating the images differently. The same can be said of text, chemical structures and other areas. In all instances new technology allows scientists to do different types of experiment than otherwise they could have.

Deep learning has ramifications across the entirety of the pharma area. With greater analytical and predictive ability, scientists can institute global, large-scale programs to better run R&D. This is changing the nature of pharmaceutical problems that currently cannot be answered.

The Rise of China

Commenting on the companies that were performing the most innovative, forward-thinking work in pharma AI at the moment, Dr. Zhavoronkov said that beyond his own company and certain other biotechs like Deep Genomics, there were a number of innovative companies in China that could not be ignored. 

General statistics buoy up these claims, with Chinese investment in biotech and drug discovery rising sharply in 2019 to $1.4 billion, compared to $125.5 million in 2017. In 2017 the Chinese government released an AI strategic plan, declaring the goal of catching up in the AI race by 2020 and becoming a world leader by 2030. 

China benefits greatly from the size of datasets created from its population, with reduced privacy laws facilitating greater access than is available in some other countries. It has been bolstered by rapid migration of experts from other parts of the world, and governmental policies which push research forward. However, a lack of core pharmaceutical skills and less intellectual property protection will ensure the catch-up is not as swift as it might otherwise be.

_____

Those are some of the ways AI will change pharma in 2020, but certainly not all: you can find more information on these and other ways in the full R&D white paper, available to download here. 

As always, I look forward to discussing these issues with you throughout the year.

Louis Smikle, CEO
Proventa International