So-called lighthouse projects are typically defined as small efforts focused on deliverables in a narrow area, designed to establish a pathway for larger enterprises down the line. But as the biopharma industry’s digital transformation continues apace, with the rapid acceptance of tools such as artificial intelligence and machine learning, certain projects promise to yield impacts on a much wider scale.
At the same time, AI expertise is being quickly diffused throughout the industry through collaborations, with numerous tech firms offering different approaches and services as drug and device makers shop around for those that best suit their needs.
In just the past few years, biopharma has built a large, interconnected network of partnerships as it aims to transform and digitize its processes to maximize their value—across everything from molecule design to clinical trial planning, as well as in supply chain, quality control and sales strategies.
In a survey released earlier this month by Optum of 500 healthcare industry leaders and professionals, the number of respondents who said their organizations had an AI implementation strategy in place had increased by nearly 88% compared to the year before.
Additionally, the survey found that organizations plan to spend about $40 million apiece on average on AI-related projects over the next five years—and half of respondents said they expect to see positive returns on AI investments in three years or less.
“It’s encouraging to see executives’ growing trust in, and adoption of, AI to make data more actionable in making the health system work better for everyone,” Optum President and COO Dan Schumacher said in a statement, although higher levels of trust in AI were seen in regards to administrative applications compared to clinical ones, and the automation of business processes was ranked higher as a priority.
Still, this sets the stage for potentially rapid adoptions in R&D and care delivery as new methods are validated and become available.
Forming collaborations and partnerships will be paramount, as many companies lack the expertise needed to make the transformation on their own in the coming years. Among medtech companies specifically, a report from Deloitte predicts that the current illness-focused system will be completely overhauled by 2040 and replaced by a proactive one that integrates data to personalize a continuum of care spanning before and after a procedure.
One of the ripest sectors for advancement is in digital pathology and assisting diagnosis. Many research projects aim to use machine learning processes to spot the patterns of diseases or conditions in images or scans. This can help ease the burden on hospital departments by screening chest X-rays, MRI scans, tissue slides or pictures of the eye.
In April 2018, the FDA approved the first medical device in the U.S. to use artificial intelligence to detect cases of diabetic retinopathy, the most common cause of vision loss among people with diabetes. Using digital images uploaded from a retinal camera, the AI—dubbed IDx-DR—can detect those with more than a mild case of the disease and refer them to a healthcare professional.
The benefits of automating healthcare and research processes are similar in other areas of medicine—not just in mining data for insight but also in sharing those insights.
Take the MELLODDY project, for example: short for Machine Learning Ledger Orchestration for Drug Discovery, the initiative hopes to share preclinical data among a network of Big Pharma companies and research partners using a blockchain-based infrastructure to protect confidentiality and proprietary information.
Meanwhile, the newest center of the NIH is working to build a universal translator for medical data, with the goal of bringing together researchers from different fields across the healthcare enterprise—on top of redefining our current definitions of disease based on the findings.
Machine learning can also spot patterns in a flood of data from multiple sources, where certain changes, no matter how small, can herald the onset of Alzheimer’s disease and dementia. Evidation Health, along with Apple and Eli Lilly, is looking to develop digital biomarkers for neurodegenerative disease by tracking people’s daily routines, device usage and changes in speech.
Using all the information provided by patients from different angles, and then feeding that data back into care delivery to potentially improve outcomes, is where machine learning can excel, and Verb Surgical—the joint venture between Verily and Johnson & Johnson—aims to use those tenets to drive a new generation of surgery.
Elsewhere in the Google/Alphabet sphere, DeepMind has developed an AI program to take on one of the most challenging problems in medical science: predicting protein folding, a mathematical problem that reaches a spectacular number of possibilities. The field once relied on human intuition to solve the puzzle. Now we’re teaching machines to follow similar instincts.
But the biggest changes will come when these methods, tools and knowledge can be made widely available, and the public-private ATOM consortium is looking to do just that. Spun out of the U.S. government’s cancer moonshot efforts, the initiative by GlaxoSmithKline, UC San Francisco and federal research laboratories hosts a series of research projects aimed at accelerating preclinical development to a timeline under one year.