Skip to content
Bindlix

Platform / Science

A design discipline for protein binders

We don’t screen millions of molecules and hope. We specify the exact surface we want to bind, design binders against it in silico, and reject everything that fails before it ever reaches the bench. The result is a small set of high-confidence candidates instead of a haystack.

The Bindlix Engine

Five-stage cascade

  1. 01

    Targeted generation

    Starting from an experimental structure (cryo-EM, crystallography) and a defined epitope, we generate de novo binder candidates using state-of-the-art design methods — hallucination- and diffusion-based. Each binder is built to engage a specific set of hotspot residues.

  2. 02

    Conformational counter-screening

    A binder is only useful if it hits the right shape. Every candidate is screened against a panel of decoy structures — other conformers, other assembly states, related isoforms — and rejected if it binds any of them. Selectivity is designed in, not discovered later.

  3. 03

    Developability filtering

    Surviving designs pass hard biophysical filters: aggregation propensity, buried unsatisfied hydrogen bonds, net charge, predicted thermal stability, and independent fold confidence. Designs that would fail in expression or formulation are removed early.

  4. 04

    Multi-objective ranking

    Candidates are scored across affinity, selectivity, stability, and structural diversity, then ranked — so the panel we carry forward is both high-quality and non-redundant.

  5. 05

    Experimental validation

    The top panel is expressed and characterised — binding, selectivity, stability, and function — closing the loop between design and reality.

Computation narrows the search. Experiments confirm it. Neither alone is enough — the platform is the loop between them.