Accelerating insights into cancer therapy development

Each year, more than 20 million people around the world are diagnosed with cancer, including nearly 2 million in the United States. Pharmaceutical companies that develop treatments for cancer are constantly searching for new strategies to create and distribute safer, more effective, and more accessible drugs as fast as possible to address these significant needs.

Currently, the research and development (R&D) process can take decades, limited by increasingly outdated legacy approaches to scientific discovery, data processing, and clinical trials. As cancer cases continue to increase across an aging population, biopharma sector leaders need to investigate new approaches to clinical research that take advantage of the latest in data science and integrated infrastructure.

Multimodal data will be a key part of this endeavor, believes Jenny Wei, PhD, senior director and head of R&D Informatics & Technology at Kite Pharma, a Gilead Company.

At the 24th Annual Bio-IT World Conference & Expo held in Boston in 2024, Dr. Wei was joined by Andras Fancsik, MD, PhD, specialist leader in Life Sciences at Deloitte Consulting LLP, to share their perspectives on the value of using multimodal data to augment traditional methods of clinical investigation.

After their presentation, entitled Building Cell Therapy in a Multimodal Scientific Ecosystem, we asked Dr. Wei to expand on some of the key concepts surrounding multimodal data and how pharmaceutical companies can leverage a rich array of data assets to enhance clinical discovery, reduce time to market, and ensure all people with cancer get the care they need.

How would you define ‘multimodal data’?

Dr. Wei: Multimodal data is simply an aggregated set of data that contains multiple data formats. In the biopharma sector, we are actively expanding the type of data we can use for exploring the development and impact of new cancer therapies. Novel data types might include imaging data, waveforms, unstructured text, a huge range of biomarker and omics data, and even voice data in the near future.

We want to use as many different types of data as possible to give us the most complete and accurate view of what’s going on inside the patient’s body in response to a therapy. Some of these data types will be more applicable to certain use cases than others, so we need to gain a better understanding of how, when, and why to use specific collections of multimodal data.

In the emerging world of chimeric antigen receptor (CAR) T-cell therapy, for example, the discovery process can be complicated. Many different data modalities are created across a wide range of business capability groups. This includes flow cytometry data, which comprises more than 60% of the active data across cell therapy research.

Flow cytometry data includes data on cell viability, proliferation, and sorting, as well as immunophenotyping. These types of data are crucial for understanding disease relapse and recurrence mechanisms and for getting the genetic engineering right to have an impact on solid tumors. We want to make as many correlations and connections between these data types as possible to unlock new discoveries in a faster and more efficient manner.

Ideally, leveraging a broader range of multimodal data will help us identify why some patients respond well to therapies and others do not.

What are the challenges of generating, aggregating, and leveraging multimodal data?

Dr. Wei: Multimodal data, by its very nature, comes from many different sources: internal labs, clinical trial sites and other business partners using a variety of assay techniques, electronic health records, and third-party data providers. It’s critical to have the right infrastructure in place to ingest and aggregate these data types.

There are a lot of logistics to manage and a lot of partners to work with—often more than 30—to bring the data into the environment. And there isn’t just one set of data: We look at patients over time, so we continually receive data over long spans of time that have to be matched to the right individual.

For many pharmaceutical companies, this aggregation work is still very manual, which can slow down the process. At Kite/Gilead, we are working on ways to accelerate integration so we can derive insights more quickly. Copying and pasting is the bane of every scientist’s existence, so we are continuing to enhance automation.

We started by taking advantage of the cloud’s ability to support the near real-time transfer of data from sources such as flow cytometer instruments. From there, we use cloud-based tools that can process these files, interact with the metadata, and standardize key elements. Using API interfaces and gateways, we are able to automatically register these data files, which dramatically reduces the manual burdens of working with hundreds of files for each experiment.

All of this happens based on governance and business rules that are built into the system, so the process can be executed efficiently and in a standardized way without the need for end users to interact with the process at all.

What should the future of multimodal data in biopharma look like? What are you most excited about?

Dr. Wei: In my particular area of focus, I’m eager to see multimodal data enabling breakthroughs like 3D cell cultures that give us incredible precision for modeling diseases and observing how individual cells interact with therapeutic agents. The ability to continuously monitor cells in real time would be game-changing for cancer research and therapy development.

As an overall industry, however, the biopharma sector needs to focus on moving away from legacy mindsets and technologies toward an ecosystem-based approach to accelerate discovery. FAIR data must be at the heart of any multimodal data ecosystem. That means data must be findable, accessible, interoperable, and reusable.

To achieve this, we envision building a cell therapy research ecosystem that is based on three pillars: cloud native, automation focused, and prediction capable.

Leaders should be thinking about how to apply these principles to the entire ecosystem, not just one step or one part of the process. The earlier we can establish those three pillars across the life cycle, the better.