How Life Sciences Companies Are Leveraging Machine Learning to Improve Speed to Market
Successful drug research and development can take several years, if not a decade or more. While many factors contribute to it being such a lengthy process, one that’s particularly notable is infrastructure. That’s because at most life sciences companies there’s a natural wedge between data science and IT teams. As a result, building the infrastructure necessary to run drug trials usually takes a lot longer than it should.
In many cases, before data scientists can begin to de-identify the clinical data they have to work with, for example, they have to get IT’s approval, order the infrastructure, deploy it in their data center, and get a system administrator to configure it in a way that works. Unfortunately, all of those tasks can add six months or more to the overall research and development (R&D) process.
The good news is that today’s life sciences companies have an opportunity to accelerate this process, and ultimately improve patient care, by using machine learning (ML) and artificial intelligence (AI. In this article, we’ll take a closer look at some of the ways in which that’s happening.
How Machine Learning Is Changing the Way Life Sciences Companies Address Clinical Trials
Machine learning has a number of different applications within life sciences. One of the most important is to correlate structured and unstructured data to provide answers to complex questions. Another is to enable predictive modeling to support the detection of conditions like sepsis.
In the life sciences, and the corresponding clinical trials specifically, we’re seeing companies de-identify patient records using cloud-based machine learning technologies so that they can run clinical trials against geographies and demographics of the data they have. That way they can correlate where the commonalities are against a patient and a clinical trial population. That, in turn, allows data scientists to focus solely on asking the right questions and building algorithms behind the scenes to support those questions, while letting the machine learning technology from public cloud vendors do all of the front-end work.
Johnson & Johnson is a great real-life example. The company uses de-identified patient data to understand how its products are performing. In doing so, the company tries to assess if people are benefiting from their products as much as they were in clinical trials and if new sides effects are emerging that physicians need to be made aware of. This has also enabled Johnson & Johnson to analyze more data then ever before, while also empowering them to respond in a more agile way to changes in development.
To that end, another important application of machine learning in the life sciences is the ability to enable much more complex and sophisticated segmentation so you can track individual behaviors across millions of users inside a patient population. This has helped drive better commercial experiences, so when drugs are ready to be tested through clinical trials, they get a quicker response and higher levels of engagement. In turn, this results in real-time feedback loops that can be used to expedite the approval process.
Practically speaking, advances like this can lead to faster time to market and ultimately cost savings. Considering that on average, it costs $4 billion to develop a new drug, anywhere you can help reduce costs is critical to profitability.
Of course, it’s not just during clinical trials that machine learning has a role to play. It’s also critical in manufacturing and providing quality assurance for new products, as well as for analyzing the growing datasets associated with real-world evidence. According to the 2018 Deloitte Real World Evidence Benchmarking Survey, 60% of life sciences companies are currently using machine learning to analyze real-world data — but almost 95% expect to use it in coming years. Other areas that could be impacted include supply chain management, lot release, compliance operations, and drug discovery, to name just a few. By leveraging machine learning services, combined with public cloud technology, all while following GxP for quality assurance, life science organizations are able to prove supply chain compliance.
If your company is ingesting personal health information (PHI) into any kind of enterprise data warehouse or public cloud, you need to take care to follow any rules set forth in regulations like HIPAA, GDPR, and GxP. That includes requirements like encryption in transit and encryption at rest, to name a few. In addition, all the tooling that data scientists use to view, monitor, and log PHI activities need to have a business association agreement (BAA) around it. That agreement is the promise that third-parties make to treat, process, and store PHI in a secure manner.
De-identification is another important consideration around PHI. There are lots of tricks of the trade and interpretation around de-identification, making it critical to align with your legal teams before running clinical trials against any patient data.
Importantly, if your goal is to go to market faster by harnessing the benefits of PHI in the cloud, the key is to use the existing healthcare reference architectures or blueprints from public cloud providers like Microsoft Azure, Google Cloud Platform, or Amazon Web Services. Those blueprints will give you a good idea of how other life sciences companies are already using PHI in the cloud. In addition, make sure to take advantage of the scale and agility that public cloud providers offer you, so you can focus more of your attention on your core competencies.
A Better Way Forward
By allowing life sciences companies to aggregate vast amounts of data, machine learning is helping to accelerate both the process by which drugs get tested and manufactured, and how they ultimately reach patients. Although we’re still in the early days of embracing machine learning, the promise of agility that it brings is very real. If you want to accelerate your drug and clinical trial research, using the public cloud will help by allowing you to spin up the infrastructure you need in a matter of minutes without having to go through IT.
One thing to remember: As you apply machine learning to PHI throughout the drug and device life cycle, security and compliance breaches can happen if you don’t take the right precautions and understand the constant change throughout the cloud industry. That could mean hiring dedicated resources to manage these critical elements, or seeking third–party support.
In the case of a partnership, the right partner will understand the unique needs of a life sciences organization looking to innovate in the cloud, helping you mitigate risk as you as you work through clinical trials and beyond.