Setting a New Speed for Life Sciences in 2024

Scaling AI Kelci Miclaus

The beginning of a new year is always ripe with opportunity as well as reflections. If I look back at the trends I believed to dominate the life sciences field last January, they centered around: 

  • Health and therapeutic access equity
  • Real World Data (RWD) driving paradigms with multimodal and causal models
  • ChatGPT and the rise of Generative AI 
  • Composable architecture implications driven by all the above

Incredible progress has been made in the journey to embed AI into every facet of life sciences organizations. In 2023, a key trigger was how to maintain speed in innovation during what could be called “pandemic recovery mode,” whereas 2024 already feels like it has been hit by a bolt of energy akin to what turned Barry Allen into a speedster in the Flash. That bolt is formed by technology advancement, strategic collaborations, and Generative AI; and it is setting a new standard for what the pace of change looks like. Because of it, we can expect to see superspeed in the following trends in the coming year.  

the flash

Initial ROI From Risk-Managed LLM Exploration and AI Governance Strategies

It was no surprise to industry veterans to see Generative AI approached by life sciences organizations with a risk-mitigation-first approach. Many organizations locked down access to tools like ChatGPT while at the same time began rapidly evaluating emerging Large Language Model (LLM) technologies to consider both opportunity but also safety, security, governance, and potential risks or limitations.  

Established data science teams in life sciences organizations recognized early this isn’t an “out with the old, in with the new” paradigm but just an inflection point of how to incorporate generative technologies into their existing analytics and machine learning (ML) practices and create more holistic enterprise data strategies.  

Pharmaceutical organizations are no stranger to regulations, and new governing body frameworks such as the EU AI Act supplemented already established regulatory guidances such as AI software-as-a-medical device (SAMD) and Good ML Practice (GMLP). And so they are versed better than most on how to plan adoption strategies in a risk-controlled manner.

This industry has rich experience around AI applications in high-risk (e.g., patient outcomes or safety considerations) scenarios and the imperative for Responsible AI practices (which has always been a requirement when it comes to health decisions). Life sciences organizations therefore have the potential to show-off their newfound speed and outpace other industries in adopting generative technologies to realize ROI in their business models, particularly in areas such as drug discovery and medical writing. 

AI for Everyone

I’ve had hundreds of conversations with both data and business leaders at top life sciences and technology orgs this year who share our Dataiku mission toward Everyday AI — systematizing the use of data to drive business value. Even as our platform has been adopted and deeply embedded for everyday use in core data and IT teams, both in global pharmaceutical organizations as well as smaller digital native biotech players, bringing AI to the business teams is still in its infancy (in the industry as a whole) and is the linchpin to realized AI impact on strategic business outcomes.  

Perhaps one of the things I am most excited to see from the Generative AI hype craze is the paradigm shift where lines of businesses are now ASKING to be a part of this revolution. They want on the speed train, especially when there is not a Jupyter notebook or S3 bucket connection as the barrier of entry.  

Life sciences organizations like Moderna and Merck doubled down last year on creating lasting AI literacy and data democratization strategies for their companies.

Whether it’s code assistants to extend the capabilities of entry-level data scientists and engineers, bench scientists and biologists guided by LLM-powered hypothesis generation, or marketing campaign managers building tailored personalized content — everyone in an organization can begin to ask questions of even the most complex documents, data, and predictive analytics models to drive value now.  

Some of the early adopter biopharmaceutical leaders have already put in place their own enterprise-wide RAG-powered chatbots with thousands of active daily users. Largely that use has been for internal alignment, document management, HR, and training but, this year, expect to see healthcare companies scale and expand to creation of knowledge bases with embedded conversational AI that drives market intelligence, pipeline development, clinical operations,  market access, and commercial engagement as a core component of their business processes.  

The Journey to Multimodality Is More Than LLMs

Few industries present the scale or complexity of data like healthcare and, as powerful as LLMs have proven to be, a core trend toward integrating diverse data modalities holistically is just starting to be seen. Whereas much of last year was predominantly focused on LLMs, we started to see some of the early foundational models build in deeper multimodal functionality. 

This year, I expect to see the journey continue by deeper development paths on progression of large models for other slices of data modalities. Namely, image processing and large vision model (LVM) applications are looking to “catch-up” to the LLM pace, with more and more digitization in areas of single-cell imaging and spatial transcriptomics, medical imaging, digital pathology that promise to unlock potential value particularly in diagnostics for precision detection, diagnosis and treatment of disease. To borrow the acronym habit around good “fill-in-the-blank” practice (GxP) frameworks, LxMs in other areas than just language may just take a front seat.

Domain-specific contextual models will also continue to be a focus along with improved development of foundational LxMs. Varied applications, costs, and utility will drive the need for a spectrum of generative technologies, and “large” will need to become relative to the delivery mechanism. For example, companies need to ask themselves now,

What will Generative AI look like in the context of edge computing for digital health apps, in-home diagnostics, or wearables tracking health metrics whose utility relies on the interoperable ecosystem across patients, providers, insurers, and drug manufacturers?

A Convergence of Life Sciences and Technology

Speaking of hardware and software … In a trend that has already taken off faster than a speeding bullet this year, technology giants such as Google, AWS, and NVIDIA are making deep investments in their services tailored to the healthcare markets and biopharmaceutical and biotech companies are hungry for it.  

These investments, much like our own investments in new Dataiku capabilities based on the needs of our industry partners, are fed by the collaborations and aligned strategies between pharma and tech. Look no further than NVIDIA introducing their BioNeMo platform cloud APIs and the partnerships with Amgen and Genentech already forming to use high performance computing and tailored biological libraries to tackle the massive challenges in AI discovered, AI designed drug development.  

Similarly, pharmaceutical giants Eli Lilly and Novartis are deepening their technology in the discovery and development space with Isomorphic (the Alphabet company whose foundation is built upon AlphaFold from Google DeepMind).

Life sciences organizations are becoming technology organizations, and it goes beyond AI as we see connected digital manufacturing and robotic lab automation becoming a reality to creating the digitized molecule to market process. One of the most challenging components of these technology investments can be the time to impact when we see AI-embedded R&D programs bring a new therapy to the market.  

If and when we reach that milestone, we also need market engagement and diagnostics ready to keep pace as well. Early evidence is already there with the recent FDA approval of BrainSee, a medical imaging algorithm that uses AI to predict a patient’s likelihood for progression from aMCI to Alzheimer's dementia within five years.

Injecting AI-Driven Insights Into Medical Affairs and Digital Health

Speaking of FDA approvals, last year we saw near record-level approval of 55 novel drugs approved to the market, spanning the entire gamut of therapeutic areas. While a lot of the excitement and focus is on how AI is transforming the discovery, development, and operations space, I’m particularly excited about how AI (both generative and existing data and ML applications) is being adopted by medical science liaisons and the medical affairs community. 

These teams are tasked with the daunting task of navigating the increasingly digital engagement channels to educate both doctors and patients about the new therapies available and provide timely insights to ensure new therapies improve patient outcomes and any potential risks are appropriately characterized and addressed. As part of the technology adoption trend, pharmaceutical organizations recognize the need to “own the digital experience” of a given treatment in order to improve delivery of the right therapy to the right patient at the right time.  

Dataiku, as a data science orchestration platform that can incorporate both established and emerging technology as well as diverse business needs, is perfectly primed to be a strategic accelerator in these innovation areas. Together with our technology and our life sciences partners, I look forward to seeing what the speed force will bring for 2024. 

You May Also Like

Advancing Healthcare With Dataiku and NVIDIA

Read More

How Can CIOs Bridge the Gap Between Modern Analytics Aspirations and Reality?

Read More

Are AI Agents the Answer to the Commodity AI Trap?

Read More

Unpacking the EU AI Act and Its Impact

Read More