It’s time to stop tiptoeing around applying Artificial Intelligence (AI) to the development and manufacturing of therapeutic drugs. If we use this tool carefully and thoughtfully, we will bring benefits to researchers, clinicians, and ultimately patients. The eventual widespread application of AI will stop the wasting of time and resources, and instead allow us to act quicker, based on insights from AI and other digital technologies.
Note that I called AI a tool. I’m not calling it the answer to everything, although it can answer a lot of previously unanswerable questions. During my 20 years working in the healthcare industry leading the development, approval, and launch of numerous drug and diagnostic products I have been able to see many opportunities realized. There is an enormous desire to progress and help bring new treatments to patients. But now, with AI at our fingertips, we’re moving ever closer.
Here are five ways I see AI helping to reduce variables and speed up breakthroughs across the pharmaceutical and Life Sciences industries:
1) AI helps understand diseases better
The many variables of early-stage drug development can cripple progress. Which drug molecule out of thousands to take forward to clinical development? Which disease site to target in a complex molecule? I believe AI and its subset, machine learning, are the technologies that will finally deliver on the promise of smarter Research and Development.
At GE Healthcare, we recently introduced a machine learning module called “Phenoglyphs” to pair with our IN Carta and IN Cell software, used to understand cells. Phenoglyphs can understand various characteristics of the cells it is analyzing to uncover phenotypic differences in a patient’s cell population. This can lead to the identification of unexpected subpopulations that never would have been detected manually. Increased efficiency in the work to understand disease results in better-targeted treatments.
2) AI could address pharma’s decades-old challenges
Clinical trial failure rates are some of the costliest practices in the healthcare industry. In the highly-promising space of immunotherapies, treatments which use the body’s immune system to fight cancer, thousands of trials are taking or due to taking place. However, it’s expected that a significant number of these will fail because the wrong patients are selected or the disease incorrectly stratified, thus costing pharmaceutical companies billions each year.
GE Healthcare and Vanderbilt University Medical Center are collaborating on making AI-powered applications and new imaging PET-tracers to make immunotherapies safer and more accessible. The idea is to combine data science, genomic, imaging, and cellular analysis capabilities. The applications will look at specific toxicity types to guide the data curation and the extraction process, and generate tangible results to help clinicians determine which participants would likely respond well to treatment.
3) AI could optimize the manufacturing process before you build the factory
We would all like to be able to look into the future, especially when you are planning to invest time and money. The difficulty of where to place bets is particularly apparent in the variable-heavy business of cultivating cell cultures for the production of biopharmaceuticals.
Artificial Intelligence is one of the technologies behind digital twins, self-learning analytical engines. GE Healthcare is currently testing digital twins of cell bioreactors that grow cell cultures. A digital twin of a bioreactor could predict how the cell culture is going to improve, any expected deviations, and so produce information to optimize the process. In the future, digital twins could be applied before facilities are built allowing biomanufacturers to optimize their processes from the start.
4) AI is transforming analytics which could help foresee disease
Alzheimer’s Disease is a neurodegenerative condition affecting millions of patients worldwide. Clinical trials for Alzheimer’s drug therapies have historically suffered from high failure rates due to heterogeneity in both patient disease manifestation and non-optimal patient selection. GE Healthcare is developing a predictive analytics digital biomarker, Cerebro, helping pharmaceutical companies improve the patient selection process for clinical trials by making it possible to combine patient data from imaging exams with clinical information such as psychometric test scores, demographics, and genetic testing.
The accuracy of the digital biomarker algorithm meant that when testing it only a relatively small number of patients needed to be included in the study. Key to success was the meticulous collection of data from various sources including patient cognitive tests, demographics and MR and PET imaging results to supplement the algorithm. This algorithm showed around 90% accuracy and has the potential to optimize Alzheimer’s clinical trials and care pathways by helping to identify the right patients for clinical trials and the right trial design.
5) AI is empowering people
In applying AI to healthcare, we know that to be successful we needed to partner, both to have increased data visibility as well as valuable expertise. GE Healthcare’s alliance with Roche Diagnostics is co-developing tools to help inform decision-making when it is most crucial, including in acute situations, such as identifying the onset of sepsis. Without prompt treatment, sepsis, a common but serious complication arising from an infection, can cause multiple organ failure.
The companies are co-developing tools to combine patient’s in-vivo and in-vitro information alongside patient records, medical best practices, and the latest research outcomes. One aim of the partnership is to create an AI-enabled “virtual collaborator” to integrate data from Electronic Medical Records with other hospital systems to provide insights into patients who are at-risk for sepsis-related deterioration. The goal is tools that will be able to help physicians make more informed, earlier, faster diagnoses, and help them determine the most appropriate, individualized treatment.
Artificial intelligence is regarded as a disruptive technology and the connotations of its application in pharma can be ominous; replacing GPs with algorithms, anyone? Yet any good scientist knows that experiments should only test one variable. At its core, AI and other digital tools are doing just that, analyzing the variables and producing data so that humans are left to make the important decisions.