Validating cBTE: p38α Hits, Potencies, and a Head-to-Head with Classical BTE
A new screening method is only useful if it finds real, potent binders — not just DNA sequences that happen to be enriched. For cellular binder trap enrichment (cBTE), we validated this two ways: biochemically, by testing p38α hits in an orthogonal enzymatic assay, and methodologically, by comparing the hits cBTE found against those found by classical, purified-protein binder trap enrichment (BTE) on the same target.
The hit count, target by target
Screening the 194-million-member library Lib051 against three targets — p38α, ACSS2, and DOCK5 — inside living oocytes, across nine parallel experiments (three bait-DNA concentrations × triplicate), produced 323 structurally unique hits above the mathematical significance threshold:
| Target | Specific hits | Notes |
|---|---|---|
| p38α | 154 | Multiple distinct chemical clusters |
| DOCK5 | 76 | Multiple distinct chemical clusters |
| ACSS2 | 20 | Multiple distinct chemical clusters |
| Non-POI / promiscuous | 70 | Shared background across targets; 3 additional singleton hits unclustered |
The presence of several independent chemical clusters per target — rather than one dominant family — is itself informative: it suggests the screen is finding genuinely diverse binding chemotypes rather than a single artifact repeatedly re-discovered.
Biochemical validation for p38α
We selected 21 p38α hits for follow-up, deliberately sampling across the data set — different physicochemical properties, different signal strengths, and both the largest structural cluster and its most dissimilar counterparts — rather than cherry-picking only the strongest apparent signals. Each was resynthesized on solid phase and tested in the ADP Quest biochemical assay (against the reference inhibitor SB203580, IC₅₀ = 87 nM).
Of the 21 compounds tested, 15 showed IC₅₀ values below 10 µM and 9 below 1 µM. The ten most potent:
| Compound | Biochemical IC50 | Cellular IC50 (TNF-α secretion) |
|---|---|---|
| VPC00628 | 25 nM | 7 ± 0.9 nM |
| VPC00257 | 0.14 µM | 46 ± 2.5 nM |
| VPC00630 | 0.17 µM | 44 ± 4.4 nM |
| VPC00251 | 0.39 µM | — |
| VPC00262 | 0.49 µM | — |
| VPC00249 | 0.51 µM | — |
| VPC00626 | 0.69 µM | — |
| VPC00261 | 0.86 µM | — |
| VPC00797 | 0.93 µM | — |
| VPC00254 | 1.9 µM | — |
The three most potent compounds — VPC00628, VPC00630, and VPC00257 — were also active in a cellular assay of TNF-α secretion in a human monocytic cell line, at single- to double-digit nanomolar potency. VPC00628 had previously been identified independently by classical BTE, where its binding mode in the p38α active site was confirmed by X-ray crystallography (Kd = 2.5 nM; PDB accession 5LAR). Its rediscovery here, by an entirely cell-based screen, is itself a form of validation: the same high-quality chemical matter is reachable whether the target is purified protein in a tube or a fusion protein expressed inside a living oocyte.
Comparing cBTE against classical BTE, head-to-head
Because p38α had already been screened against a related library by classical BTE — the purified-protein, in vitro version of the same trapping chemistry — it offered a direct opportunity to ask whether moving the screen into a living cell changes what gets found. Comparing the hit sets, we observed good overlap between the two methods, with all chemical clusters identified by cBTE also represented among the classical BTE hits. This agreement is a meaningful validation point precisely because p38α is known to behave well and give few screening artifacts in a conventional in vitro screen — it is a fair, relatively “easy” benchmark case for asking whether the more complex cellular method reproduces a trusted result before extending it to harder targets.
Reproducibility across replicate experiments
Beyond the BTE comparison, internal reproducibility was also assessed directly: across the three bait-DNA concentrations tested per target, Venn-diagram analysis showed that 30–50% of hits for a given target were observed in all three independent experiments, with the remainder appearing in one or two of three — the pattern expected from real signal combined with the stochastic sampling variation inherent to screening a library this large, rather than from an unstable or unreliable assay.
Source: Petersen, L. K.; Blakskjær, P.; Christensen, A. B.; Andersen, J.; Folkesson, C. G.; Kristensen, O.; Andersen, C.; Alzu, A.; Sløk, F. A.; Madsen, D.; Azevedo, C.; Micco, I.; Hansen, N. J. V. “Screening of DNA-Encoded Small Molecule Libraries inside a Living Cell.” J. Am. Chem. Soc. 2021, 143, 2751–2756. DOI: 10.1021/jacs.0c09213. Compound data cross-referenced against Petersen, L. K. et al. “Novel p38α MAP kinase inhibitors identified from yoctoReactor DNA-encoded small molecule library.” Med. Chem. Commun. 2016, 7, 1332–1339.