Building an Agile Approach to Antibody Discovery

September 16, 2021 | When the senior leaders at Eli Lilly and Company heard Helen Li’s plan for the pharma’s new Biologica platform for harvesting company institutional knowledge and building an analytic engine to suggest high quality antibody molecules for early discovery pipeline, they had one challenge: do it faster. And so, Li, Head of IT at Lilly Biotechnology Center, launched an accelerated, agile development plan to build a platform they envisioned to house all experimental data, that feed into in silico modeling, a platform to scale up analytics to empower quality by design.

Historically, scientists analyze data in silos, sharing learnings across projects is sparse. Data were not FAIR (Findable, Accessible, Interoperable, Reusable).

Lilly invested in automation to increase the throughput of antibody harvest and screening. Li’s team developed an IT system that automated and standardized experimental data and developed analytics to sift through millions of data points so that researchers could make quality molecule selections.

Li will be sharing the Biologica platform at the upcoming Bio-IT World Conference & Expo held both online and in person, September 20-22. She gave Stan Gloss, founding partner at BioTeam, a preview.

Stan Gloss: How did you go about developing the Biologica platform?

Helen Li: The goal of Biologica is to understand the underlying patterns between molecule structure and its activity and property that was not obvious and be able to find and suggest the “ideal” molecule to move to clinic with higher success rate to the market.

To accelerate the effort, we tackle the problem on three fronts:

1. Integrated LIMS system: we deployed a system well integrated into the workflow, so we could capture data contemptuously, and lowered the barrier for data getting into our system.
2. Standardize and centralize experimental data to ensure high data quality to build machine learning model
3. Integrate and innovate on analytics to find patterns in the data that could derive actionable insights and move the needle of quality by design.

What was the relationship like between your team and the end users?

I think we're very fortunate. It started with good alignment of senior leadership between business and Research IO, Ramesh Durvasula [VP of Research IT and Informatics], they are all invested in the success of the program.

At the execution level, our team were embedded in scientific working groups. We worked with scientists in lock step, explored the possibilities, standardized the processes, and designed the solutions together. With any large transformation, change management is always the key that makes or breaks it. Rallying scientists around the common goal with shared consciousness and having them as partners are proven as the key to our success.

That's terrific. The integration of multidisciplinary teams is such an indicator of success.

Yes. It took a village to build such an elaborate platform. It is essential to have people understand the science, then translate, simplify, and design the technology solution. Our team worked as catalysts among many scientific working groups to foster collaborative thinking, pushing the boundaries, navigating unknown territory, and prototyping together with scientists. It is a very fulfilling experience for us. Now when I go to meetings and seminars, I see screenshots from our system used in the presentations. To me, that's the ultimate success. The system is being used for decision-making by scientists when they do their work.

You talked about the IT people learning the science. Did you also find in the process that the scientists learned something about IT?

Absolutely. To many of our collaborators, this is their first IT project. We took them along the journey of IT agile development. We introduced the concept of MVP—Minimal Viable Product. How did we do that? At first, people worried that they would get an inferior product that does not meet their needs. Once they see that they could have access to prototypes faster, and that this is a mechanism to help us to iteratively understand how to unite science and technology better, they loved it. We release the product updates every two weeks; scientists do get value from every single release. Now, they call it Minimal Valuable Product.

It's amazing what you can accomplish when you respect each other's domains and have interest in understanding each other. I think that really opens to helping that along. Was Biologica all homegrown? Did you make a buy-versus-build decision or is it built on something?

Biologica is the umbrella platform. It has a lot of components. For the integrated LIMS, we licensed a vendor platform, integrated, configured, and customized to our processes. In the cases of in silico and asset data management, we had hybrid solution, license software when there are mature components in the market and build our own when there were gaps. We used our enterprise platform to house the central data management system.

As pharmaceutical companies realize that their data is an asset, it’s no longer the compound library, it’s the structured data that they have. It is your competitive advantage and outsourcing that to anybody is not probably a good idea.

It is very true that high quality data are incredibly valuable. Not only it is used to inform the immediate project decisions, but cumulatively it helps us understand the broader patterns to design high quality molecules. The challenge in the field is that there are not enough high-quality data.

I am not sure how well a complete outsourcing model would work; some company is experimenting that. Instead, we have many co-development efforts, where we actively collaborated with academic labs and biotech startups to collaboratively solve the problem together.

So, what has been the biggest result and outcome from Biologica? What's the experience of the end users who get to use this new tool?

We have several capabilities that are game changing for us: with investment in automation, we increased the throughput 10-fold, and with low barrier of data entry, our data warehouse grew exponentially. The first two are the steppingstones for the most important capabilities: using in silico models at the very early stage to sifting high quality molecules for further experimental characterization, replacing some of the early, time consuming and expensive lab experiments.

The most significant accomplishment is that the Biologica system was developed just in-time to supporting discovery of world’s first COVID-19 neutralizing antibody treatment, from discovery to First Human Dose in merely three months.

That's terrific. What's the biggest lesson learned for you? Was there anything you would have changed based on your experience developing Biologica?

The biggest lesson we learn is that when we focus, we can go very fast. Our patient is waiting; it is our responsibility to innovate fast and execute with a sense of urgency.

In hindsight, I am glad that our executives asked us to accelerate, and we managed to deliver upon that request. We were just in-time to take on the fight against a pandemic, which we did not see it coming three years ago.

It was not without challenges as we try to parallelize the development workstreams to accelerate. There were a lot of moving pieces going at the same time, it was quite an effort to make sure they would eventually all fit together seamlessly in the end. Building trust, fostering open collaboration, encouraging healthy debates, allowing trial and errors are the skills we learned to be TEAM LILLY.