How to Design Antibodies
2 days ago
- #computational-biology
- #AI-tools
- #antibody-design
- AI-based tools like Nabla Bio, Chai Discovery, Latent Labs, and Isomorphic Labs now enable computational antibody design with high success rates.
- Antibodies are versatile Y-shaped proteins used in medicines (e.g., Humira) and diagnostics (e.g., $1 COVID tests).
- Traditional antibody discovery involved screening billions of candidates, but AI tools like BindCraft reduce this to tens of attempts.
- Antibody fragments like VHH (nanobodies) are easier to design due to their smaller size and can be expressed in bacteria.
- Key metrics in binder design include dissociation constant (Kd), with picomolar (pM) and nanomolar (nM) affinities being ideal for drugs.
- BoltzGen is a leading open-source tool for antibody design, achieving sub-micromolar binders in most cases.
- The antibody design process involves five steps: target selection, structure preparation, design campaign, candidate filtering, and experimental validation.
- Tools like AlphaFold 3 and PyMOL are used for structure prediction and trimming to optimize design campaigns.
- Ariax is a user-friendly platform for running BoltzGen and BindCraft, with costs scaling based on the number of designs and target size.
- Experimental validation via Adaptyv Bio tests binding affinity, with success rates varying by target and tool (e.g., ~25% for BindCraft).
- Beyond binding affinity, specificity, immunogenicity, and half-life are critical for therapeutic antibodies.
- Future advancements may include multi-target binders, pH-responsive designs, and reduced immunogenicity through AI-guided optimization.