Science explainer

How TCELERATOR screens 10 million single-cell interactions in one run

Conventional discovery tests binders one plate at a time: a few hundred candidates per run, weeks per cycle, months to a lead. TCELERATOR™ pairs one T cell with one target cell in each of millions of droplets, reads what the T cell does, and returns every hit as a sequence with a ranked functional readout.

One T cell, one target cell, one droplet

The platform makes water-in-oil droplets at 6,400 per second, and loads them in an ordered way rather than at random. That ordering is the core of the technology. In a conventional droplet system, cells arrive at random (Poisson loading), so most droplets are empty or carry the wrong number of cells. Raising the cell concentration to get more pairs also raises the number of droplets with several cells in them, which creates false positives.

Ordered loading removes that trade-off. In a TCELERATOR run, more than 95% of T cells are alone in their droplet, 85% of those singlet T cells are paired with at least one target cell, and fewer than 5% of droplets are multiplets. At the same cell concentration, a Poisson-loaded system gives about 60% singlets, under 30% paired, and about 45% multiplets. The result is roughly 10 million usable one-to-one pairs per hour of droplet making, against about 100,000 for a Poisson system [1].

Four steps from cells to sequence

• Pair. One T cell and one target cell per droplet.
• Incubate. The T cell either recognizes its partner or it does not. Activation, cytokine secretion and killing are read inside the droplet, with incubation of up to 12 hours on the current generation of the platform.
• Sort. Droplets carrying the functional signal are sorted out. Negatives are sorted away.
• Sequence. The sorted cells are sequenced to recover the binder or the T-cell receptor behind each hit.

What a run measures

The output is a sequence with a functional ranking, not a binary binding call. The readout can be killing, cytokine secretion or an activation signal, and the scale is millions of interactions per run against hundreds in a plate assay.

For binder and T-cell engager libraries, the library is expressed by the target cells, one design per droplet, paired with a reporter T cell. Because secretion inside the droplet is proportional to incubation time, a stronger signal at a fixed time means more functional activity, which is what lets the platform rank thousands of designs rather than just sort them into yes and no. Target expression level is controlled in the droplet, and T-cell engager leads are serially titrated over the course of the experiment.

Precision and reproducibility

• Rare events are recovered at 1 in 10,000 in repeat blend-in runs (100 positive cells mixed into 1 million).
• More than 80% of post-sort clones are confirmed by bulk killing assays, run after run (range 80 to 100%), with gating tuned to balance false positives and false negatives [1].

Two jobs, one engine

TCELERATOR runs two kinds of experiment. For T-cell engager lead testing, thousands of TCE designs (binders, CD3 arms, formats) are secreted against tumor cells and read on reporter T-cell activation and killing, so that leads are selected on activity rather than on binding. This is how the small cell lung cancer and ovarian programs choose their leads.

For TCR discovery, patient T cells are paired with antigen-presenting target cells, responders are captured and their TCRs sequenced. Tumor antigen-specific T cells are rarer than 1 in 10,000 in blood, so this is where precision matters most. One run can recover TCRs for up to 50 antigens, and native antigen-specific TCR sequences come back in about 1 week instead of 6 months with plate-based methods [1, 2]. This feeds the TCR-based T-cell engager program in hepatocellular carcinoma.

Where the platform is going

The second generation (2023) delivered 10 million paired interactions per run and TCR discovery. The current third generation is designed for multispecific TCE screening, with activation, secretion and killing readouts, up to 12 hours in droplets and ranking of positive reads. The fourth generation, in development for mid-2027, extends to functional transcriptome profiling, ultra-high throughput B-cell functional screening for antibody discovery, and ranking of both positive and negative reads.

Working with us

Partners engage in three ways: a pilot screen, where a partner brings its binders or library and receives ranked functional hits while keeping all rights; a discovery collaboration, where the partner nominates targets and ShennonBio discovers T-cell engagers on TCELERATOR; and a data partnership, where design-test loops return functional labels for model training. Contact us to arrange a technical walk-through.

Related reading

• Why function beats binding affinity in T-cell engager discovery
• Why a trispecific T-cell engager for small cell lung cancer
• Why intracellular targets need TCR-based T-cell engagers
• All science explainers

References

1. ShennonBio TCELERATOR technical presentation, Oct 2026. Rare-event, flow cytometry and signal data are available under a confidentiality agreement.

2. Li K, et al. J Immunother Cancer 2025;13(Suppl 2):A383 (SITC 2025).