CD ComputaBio applies structure-based and AI-driven approaches to discover and optimize molecular glues—small molecules that stabilize otherwise transient protein–protein interfaces. We combine ternary complex modeling, interface hot-spot mapping, and glue-like degrader screening with targeted wet-lab validation to accelerate proximity-based therapeutics.
From target–ligase selection to glue chemotype prioritization, we cover the computational and early experimental steps.
Build and refine ternary complexes using Rosetta, AlphaFold2-multimer, and molecular dynamics to reveal glue-stabilized interfaces.
Identify surface cavities and hot-spot residues that can be bridged by a small molecule, using computational alanine scanning and MD-based energy decomposition.
Screen large libraries (ZINC, Enamine, proprietary) for compounds that stabilize the ternary complex, using ensemble docking and pharmacophore matching.
For targets where a bifunctional degrader is preferred, we design linker-optimized PROTACs and evaluate glue-like ternary complex formation.
We coordinate with wet-lab partners for binding affinity, ternary complex detection, and cellular degradation assays to validate computational glue candidates.
Use generative models and reinforcement learning to propose novel glue scaffolds with desired ternary complex stabilization.
We use AI-based folding and docking tools to build initial ternary models, then refine with explicit-solvent MD (AMBER/CHARMM) to capture glue-induced conformational changes.
Graph neural networks and transformer models predict glue-binding hotspots and rank compound libraries by their likelihood to stabilize a given protein–protein interface.
Each computational cycle is followed by biophysical or cellular assays. Results are used to retrain scoring functions and prioritize the next round of glue candidates.
We analyze your protein of interest and candidate E3 ligases to assess interface compatibility, available structures, and glueability.
Generate initial ternary models using AI docking and Rosetta, then refine with MD to identify stable glue-binding pockets and key interaction hotspots.
Computational alanine scanning and energy decomposition highlight residues that contribute most to ternary complex stability, guiding glue design.
Screen compound libraries or run generative models to propose glue candidates, followed by ensemble docking and MM/GBSA rescoring.
Top candidates are tested via biophysical and cellular assays. Data feeds back into the next design cycle to improve potency and selectivity.
| Biological Question | Recommended Approach | Primary Output | Downstream Decision Enabled |
|---|---|---|---|
| Can a glue stabilize a weak kinase–E3 interface? | Ternary modeling + MD + hot-spot mapping | Interface energy, glueable pocket score | Prioritize E3 ligase and glue chemotype |
| Which warhead can degrade a previously undruggable protein? | Generative glue design + virtual screening | Ranked glue candidates with synthetic routes | Select compounds for synthesis and SPR |
| How to optimize a PROTAC linker for ternary complex formation? | Linker modeling + ternary complex simulation | Linker length/geometry vs. cooperativity | Choose linker for PROTAC synthesis |
| Can a molecular glue achieve selectivity across homologous proteins? | Comparative ternary modeling + selectivity profiling | Selectivity scores across paralogs | Advance selective glue series |
| Is a natural product glue's binding mode consistent with SAR? | Reverse engineering + MD refinement | 3D binding hypothesis, key contacts | Guide medicinal chemistry optimization |
Input: Kinase domain + E3 ligase. Output: Ternary model revealing a novel glueable pocket; virtual screen identified 12 candidates, 3 validated by SPR (KD < 1 µM).
Input: TF + E3 ligase. Output: De novo glue scaffolds generated by generative models, showing favorable ternary complex stability and predicted cellular degradation.
Input: BRD4–E3 complex + linker series. Output: Ternary complex models ranking linker lengths, predicting optimal cooperativity and degradation efficiency.
Input: Natural product + target–ligase pair. Output: Reverse-engineered binding mode, identifying key contacts for SAR and analog design.
Input: Two homologous proteins + E3 ligase. Output: Comparative ternary models revealing selectivity determinants; glue candidates predicted to spare the paralog.
Input: Target with surface cysteine + E3 ligase. Output: Covalent glue candidates that form a permanent bridge, with predicted ternary complex stability and selectivity.
Molecular glues are notoriously difficult to discover by traditional screening because they stabilize transient, low-affinity protein–protein interfaces. Computational modeling allows you to explore ternary complex landscapes, identify glueable pockets, and prioritize chemotypes before synthesis—reducing the experimental burden and increasing the probability of finding a functional glue.
Yes. We use AI-based structure prediction to generate initial ternary models, then refine with molecular dynamics. If you have SAXS, cross-linking MS, or HDX data, we incorporate those as constraints.
We routinely model common E3 ligase families. If you have a less common E3 ligase, we can build homology models or use AI prediction to assess feasibility.
We coordinate with partner CROs for binding affinity, ternary complex detection, and cellular degradation assays. We can also work with your internal lab and provide detailed protocols and data analysis support.
A standard project (target–ligase pair, ternary modeling, virtual screening, and top candidate validation) takes 6–10 weeks, depending on the complexity and availability of structural data.
Absolutely. We can start from your existing molecule, build a ternary complex model, and use MD and free-energy calculations to guide linker or warhead optimization for improved cooperativity and selectivity.
Submit your project details below, and our team will respond within 24 hours.
Talk to our technical team about your project!
I Want To Talk
