DNA nanoswitch screening finds molecules that change how proteins interact
A platform called FIND uses DNA nanoswitches, gel electrophoresis and sequencing to screen billions of barcoded compounds for their effect on protein-protein interactions, rather than simple target binding. Published in Science on 5 November, the work identifies a nanobody that stabilises the interaction between the SARS-CoV-2 spike protein and ACE2, and two new small-molecule PPARγ agonists, using standard laboratory techniques.
A functional readout built on familiar methods
Most pooled screens of encoded compound libraries identify only molecules that bind a target. FIND (Functional Interaction-based Nanoswitch Discovery) asks a different question: does a compound promote or stabilise an interaction between two proteins? Reported in Science on 5 November 2026 by researchers at the Wyss Institute at Harvard University, Boston Children’s Hospital and Harvard Medical School (HMS), the platform answers it with techniques familiar to any laboratory: agarose gel electrophoresis and next-generation sequencing, applied to reactions of typically around 10 µl and resolved in a gel run of roughly 100 minutes.
Each nanoswitch is a linearised DNA scaffold carrying a target protein, a partner protein and a barcoded library member. If a compound promotes the interaction, the scaffold forms a loop, which migrates more slowly through the gel than the linear form. The looped and unlooped bands are extracted and sequenced, and each barcode’s enrichment ratio shows which compounds drive looping.
“Our idea was to build each experiment onto a single DNA molecule, so that many experiments could run in parallel in one tube,” said senior author Dr Wesley Wong, an Associate Faculty member at the Wyss Institute, Associate Professor at HMS, and Investigator at Boston Children’s Hospital.
Validation in model and receptor systems
In a DNA stem-loop model system with 4096 toehold sequences, the perfectly complementary sequence showed an enrichment ratio of 18 (P = 1.1 × 10-⁹). The effective local concentration between attached partners was 5.1 µM, allowing weak, low-micromolar binders to be characterised.
The first biological test used the β2-adrenergic receptor (β2AR), a G protein-coupled receptor, with a G protein-mimicking nanobody. Among 1024 DNA-barcoded sequences, the one bound to the partial agonist TS218 was the most highly enriched (P = 1.4 × 10-⁷), with a half-maximal effective concentration of 1.1 µM for looping.
A spike-stabilising nanobody
The team then screened a naïve mRNA-display library of about 20 billion nanobodies for enhancers of the interaction between ACE2 and the SARS-CoV-2 spike trimer. One nanobody, Nb1, stood out (P = 9.0 × 10-⁹), although it was not the most abundant in bead-based enrichment. Nb1 reduced the apparent off-rate of the ACE2 and receptor-binding domain interaction roughly 350-fold, and increased pseudovirus infectivity dose-dependently in ACE2-expressing HEK293 cells. Nb4, which did not enhance binding, did not increase infectivity. Screening about 13 million Nb1 variants identified nanobodies with higher and lower predicted activity, and all 10 tested behaved as predicted.
New PPARγ agonists from 400 million compounds
For small molecules, the team screened a 400-million-member DNA-encoded library against the PPARγ ligand-binding domain and a CBP-1 peptide after one round of bead-based prescreening. The two top library members, Ligands A and B, were not the most abundant by bead-based enrichment. Both increased PPARγ and CBP-1 association in a time-resolved Förster resonance energy transfer (TR-FRET) assay, with EC50 values of 66 nM and 310 nM, and increased lipid accumulation in a cellular assay. Their structures differ from rosiglitazone.
“We can directly search for molecules that have the effect we want on a disease-related protein-protein interaction,” said co-first author Dr Clinton Hansen, Instructor at HMS.
Limitations
The authors acknowledge that follow-up work is needed. “Mechanistic assignment generally requires additional biochemical characterisation beyond the initial interaction outcome-based screen,” they write. For Nb1, the receptor-binding domain appears sufficient for enhanced ACE2 binding, but the authors note that “other contributions from other regions in the context of the full trimer cannot be excluded.”
The authors conclude that FIND provides “a powerful and accessible research tool that should accelerate the discovery and characterisation of novel therapeutics to treat diseases.” In the Abstract, they note that “by enabling rapid discovery of molecular modulators, FIND could lead to new therapeutics across many conditions.” In an accompanying press statement it is noted that the technology has been licensed to a newly formed startup cofounded by members of the research team.
Reference:
Hansen, C. H., Lipstein, M., Erlandson, S. C., et al. (2026). Discovering molecular modulators with nanoswitch screening. Science, 394, eadt7185. https://doi.org/10.1126/science.adt7185