AI protein-fold prediction speeds molecular glue discovery: new degraders targeting VAV1 emerge
(Hanfeng Lin et al., Nature Communications, 2026; DOI: 10.1038/s41467-026-77657-z)
Key points
- What molecular glues are: small molecules that "stick" a disease-causing protein to the cell's own E3 ubiquitin ligase, so that the cell's degradation machinery clears the target protein — in effect giving the cell a "targeted trash can."
- The role of AI: the team combined AI protein-fold prediction with high-throughput proteomics to screen, from scratch and without an experimentally solved structure, molecular-glue degraders targeting VAV1 (a key protein in immune signaling), greatly speeding up early drug discovery.
- A new finding: the E3 ligase cereblon can also recognize a previously unknown binding conformation — meaning the range of proteins that can be degraded in future is larger than imagined.
- Optimization results: through medicinal chemistry that introduced a chlorine atom and reduced molecular flexibility, the team obtained the compound NGT-201-18, which has markedly higher degradation potency and forms a more stable protein complex.
- Validation in human cells: in primary human T cells, NGT-201-18 significantly lowered VAV1 levels and suppressed T-cell activation — opening a new therapeutic path for autoimmune disease, inflammatory disease and cancers that lack targeted drugs.
- A caution: some compounds also degraded another protein, LIMD1, as an off-target effect, underscoring the need for comprehensive proteomic assessment in drug development.
Source: read the original. This text is a summary for reference only.
