AI-Guided Protein Design

AI-Guided Protein Design

A multi-volume portfolio exploring how modern generative-design, sequence-design, structure-prediction, and molecular-simulation tools can be assembled into realistic computational protein-design campaigns.

The series emphasizes campaign architecture, model handoffs, candidate triage, structural validation, and decision-making under uncertainty rather than treating any individual model output as a final answer.

Campaign Architecture

These introductory pages establish the shared workflow used across the series:

• Overview of Protein Engineering

• Target Selection and Screening Funnel

• Computational Design Workflow

---

Volume 1: De Novo Mini-Binder Design Against PD-L1

An end-to-end computational campaign for designing compact proteins against the PD-1-binding surface of PD-L1.

The workflow combines RFdiffusion backbone generation, ProteinMPNN sequence design, ESMFold and Boltz-2 structure prediction, interface analysis, molecular dynamics, and consensus-based candidate prioritization.

Pages

1. De Novo Mini-Binder Design

2. Parameter Sensitivity

3. Interaction Fingerprinting

4. Molecular Dynamics Simulation

5. Independent Complex Assessment with Boltz-2

---

Volume 2: De Novo VHH Design Against PD-L1

A second design campaign extending the workflow from compact de novo proteins to single-domain antibody scaffolds.

This volume will examine how scaffold constraints, complementarity-determining region geometry, sequence design, developability, and interface validation change when the designed binder is a VHH.

Status: In development