From Emulsion to Sequencer: The Mechanics of Binder Trap Enrichment Hit-Calling

Why compartmentalize at all

Once an oocyte is broken open, everything that was inside it — the protein of interest (POI) fused to its prey tag, the bait-DNA bound to that prey, library members bound (or not bound) to the POI, and a large excess of endogenous cell protein — is now floating free in one tube. Without some way to preserve which library molecule was physically associated with which POI molecule at the moment of lysis, that information is lost the instant everything is diluted together. BTE preserves it by trapping each POI complex, with whatever library member was bound to it, inside its own tiny compartment.

The workflow, step by step

High-factor dilution. The oocyte lysate supernatant is diluted substantially into a buffer containing T4 DNA ligase before anything else happens. Dilution matters for two reasons: it reduces the chance that two unrelated POI molecules (and their bound library members) end up in the same droplet later, and — because off-rate, not just binding affinity, governs whether a bound library member stays bound long enough to be trapped — it sets the window in which weakly-held binders can dissociate before compartmentalization locks in the result.

Emulsification. Oil and surfactant are added, and the mixture is emulsified by shaking, creating a water-in-oil emulsion with far more droplets than there are POI molecules. Because POI molecules are now rare relative to droplets, each POI (with its prey, bait-DNA, and whatever library member is bound to it) is very likely to end up alone in its own droplet, physically isolated from every other POI complex and from library members that are not bound to anything.

Ligation inside the droplet. Adding Mg²⁺, the obligate cofactor for T4 DNA ligase, activates ligation only now, inside the sealed droplets. If both the bait-DNA (tagging the POI via the prey) and a library member’s DNA tag are present in the same droplet — which is the expected case exactly when that library member was genuinely bound to the POI — the ligase joins them into a single physically-linked DNA fragment. A library member that was never bound to the POI can still land in a droplet with a POI complex by chance, but this is a much rarer, purely statistical event rather than the norm.

PCR amplification with a built-in filter. After ligation, DNA is recovered from the emulsion and PCR amplified. Critically, only the ligated bait-DNA/library-DNA fusion carries the complete pair of priming sites needed for exponential amplification — so unligated library DNA (the bulk of the non-binding population) is not co-amplified into the sequencing pool at anything like the same efficiency as true trapped complexes.

Sequencing, decoding, counting. The amplified PCR product is sequenced, the DNA barcodes are translated in silico back into the small molecules they encode, and occurrences of each compound are counted. Because true binders are consistently and repeatedly trapped while non-binders only enter a droplet by chance, POI binders appear as high-frequency events against a background of low-frequency random noise — the raw signal from which hits are called.

Calling a hit: the mathematical threshold

Rather than picking an arbitrary observation-count cutoff by eye, hits are called against a mathematical threshold derived from the statistics of the background noise distribution itself. Library members with observation counts above this threshold are retained for further analysis; those below it are treated as indistinguishable from random co-compartmentalization events. This is what keeps a 194-million-member library screen from being swamped by chance associations — with that many possible library members and a finite number of droplets, some non-binders will always land in a droplet with a POI molecule purely by chance, and the threshold is the statistical guardrail against calling those noise events hits.

How much bait-DNA to use

We ran each target at three bait-DNA doses — 30, 100, and 250 fmol per oocyte — each in triplicate, specifically to characterize how labeling density affects data quality. The effect on raw signal strength was minor, but injecting less bait-DNA consistently produced cleaner data, with fewer promiscuous, non-target-specific (“non-POI”) hits. This is a useful practical dial: more label is not automatically better once the labeling chemistry is working, and a lower dose can trade a small amount of signal for a meaningful gain in specificity.

How we know the signal is reproducible

Two independent checks back up the reliability of the hit list. First, across the three bait-DNA concentrations for a given target, 30–50% of hits were observed in all three independent experiments — solid reproducibility given that some variation between replicates is expected simply from stochastic sampling of a very large, complex mixture. Second, plotting the observation count of each hit in one experiment against its observation count in a second, independent experiment shows a clear, consistent trend: hits with high observation counts in one experiment reliably also show high counts in the other, while hits near the noise floor show more scatter — and the non-target-specific (non-POI) hits cluster almost entirely at low observation counts in both experiments, exactly where background noise is expected to sit.