Hello everyone, I’m Ben. It’s nice to virtually meet you—bet you’re not tired of that phrase yet! I hope everyone’s enjoying sunny San Diego. I’d like to start by thanking Dynamic Devices for giving me the opportunity to present today.

I also want to thank all of you for attending this virtual presentation. I know it might bring back some flashbacks we’d rather forget, but here we are. Today, I’ll be presenting on how the Lynx enables scalable protein expression and high-throughput mass spec characterization.

A little about myself—my name is Ben (as you now know). I work at Bristol Myers Squibb (BMS) as a Senior Automation/Full Stack Engineer. I’ve been with BMS for around three years and am part of the Protein Sciences organization here in San Diego, within the Discovery Biotherapeutics department. I’m responsible for developing high-throughput workflows using robotic liquid handlers like the Lynx, and I also build software applications to support our scientists and help accelerate their work.

When I joined, many of our processes were heavily manual. Over time, we’ve been able to automate most of them—and without the Lynx, that wouldn’t have been possible.

Now, onto the fun stuff. As you might guess from the border on my slides, I really like food. That’s actually a big part of why I got into automation in the first place. My motto is: Work Smarter, Snack Harder. Why work harder when you can work smarter—and use the extra time to eat?

I’ll start by giving an internal process overview. At BMS, we’re heavily focused on leveraging AI and ML to help us make informed decisions. We follow a cycle called Inform–Design–Produce–Test–Inform. We use historical data to guide our engineers, who then design proteins. Those designs go into a production cycle, which continues through characterization to ensure we’re making what we think we are. Then the data feeds back into the models and scientists to improve the next round.

You’ll notice the heart of this cycle is centered around ML and the scientists themselves. Production is a critical part—it provides the materials needed for tests and assays. Everything you see highlighted in purple in the slides will be discussed in more detail today. And to feed this data-hungry system we call AI, automation is essential—it saves time and reduces error.

Traditionally, liquid handlers are highly specialized: they either produce many samples or large sample volumes (e.g., 4 mL, 25 mL; or 96, 384 wells). But not both. That’s where the Lynx excels, and I’ll dive into that in upcoming slides.

Our Stable Pool Production Process showcases the versatility of the Lynx. I’ll go into more detail about each purple-highlighted step, but generally, production begins with a chemical transfection—spiking DNA with a transfection reagent into a host line. Unlike transient transfection, we apply a selection condition afterward to eliminate low producers. This requires a media exchange.

Over the course of expression, depending on the cell line, we may add booster or feed. These stably integrated cells can produce continuously until we terminate them. To scale up quickly, we move these cells through different formats—for example, from 96-well (1 mL) to 24-well (4 mL), and then to DeepWell 6 or C-50s (25 mL or more). Cells hit critical density at around 3–4 days, and we harvest and analyze supernatant twice a week.

That harvested supernatant goes through affinity purification, followed by any sample modifications the requester needs. Often, the elution buffer used in purification isn’t assay-compatible, so we perform buffer exchange. Once that’s done, we verify the output through characterization. Since we’re working with stable cell pools, steps 6 and 7 can be repeated endlessly. These cells can be frozen and resuspended later.

Here’s our transfection process on the Lynx LM1800, which is the largest chassis available from Dynamic Devices at the time. We use a dual-arm setup: one SV (standard volume) arm and one VVP (Variable Volume Pipette) arm.

This setup supports many plate formats—from 96-well all the way up to DeepWell 6 or Centrifuge 50 (C-50) tubes. It can process 16 plates in one run. In this video (slide), both arms operate simultaneously—crucial for maintaining precise incubation timing during transfection. While one arm spikes cells at the exact time needed, the other prepares DNA plates. Their offsets are programmed to ensure they never collide.

We use this primarily for 24-well DeepWell transfections (~4 mL per well). A single run produces 384 samples. For an 8-hour day, assuming each run takes an hour, we can make ~3,000 unique molecules daily—1.2 mg for multispecifics and 3 mg for mAbs post-purification.

For screening campaigns that require more breadth but less material, we switch to DeepWell 96 format, yielding up to 12,000 molecules per day. Conversely, for fewer molecules but more material, we switch to C-50 or DeepWell 6 formats—producing up to 15 mg per well for mAbs. This workflow is used for both Hard Transient and Stable Production processes.

Let’s talk about why the Lynx is ideal for stable cell pool production. Stable cells double at different rates, making stamping heads impractical—they can’t adjust volume per well. Using a single pipette for normalization in a 24-well plate would mean repeatedly returning to each well. That’s time-consuming.

The VVP head from Dynamic Devices, part of the Lynx, solves this. It features 96 individually tunable channels, so you can aspirate and dispense different volumes into each well as needed. For instance, in a 24-well plate, four tips can fit per well, and each channel can handle a different volume. In some cases, you might need to aspirate 1.4 mL in one well and 3.8 mL in another—totally doable.

Another powerful feature is the ability to selectively activate pipette tips. For example, when filling C-50 tubes, we can fill up to six 50 mL tubes with 20 mL each—or even go up to 120 mL total across the rack—using variable volumes per channel.

During scale-up and normalization, instead of discarding excess cells to reach target density, we dilute the pool into a new plate with pre-filled media volumes. The VVP head aspirates only from selected wells, and dispenses into new plate formats (like 24 to 6-well). The format affects throughput, since pipette tips need to refill more often when changing plate types.

Once harvested, samples go through affinity purification. If the elution buffer is assay-incompatible, we use filter plates and Centrifuge Tubes for buffer exchange. Here, the VVP head’s pressure sensors detect liquid height, letting us calculate volume. That allows us to refill each well back to the target volume regardless of sample concentration, ensuring uniformity after spin cycles. This process takes ~15 minutes for up to 8 plates.