From Protein Structure to Drug Candidate

The Structure-Guided Discovery Workflow

Modern drug discovery increasingly relies on structural information to guide molecular design decisions. Whether starting from an experimental structure or a computational prediction, the workflow follows a logical progression from structural understanding to candidate selection.

Step 1: Obtaining the Structure

The workflow begins with a molecular structure. This may come from experimental methods (X-ray crystallography, cryo-EM, NMR) or computational prediction (AI-based structure prediction). Each source has different characteristics:

• Experimental structures provide direct observation but may have resolution limitations, missing regions, or represent only one conformational state
• Predicted structures can be generated for any sequence but carry uncertainty that must be assessed before downstream use

Step 2: Structure Quality Assessment

Before using any structure for discovery, its quality and reliability must be evaluated. For predicted structures, this includes examining per-residue confidence scores, identifying regions of low confidence, and understanding where the model may be unreliable for binding-site analysis.

Step 3: Binding Site Detection

Computational methods scan the structure surface to identify potential binding sites—pockets, cavities, and surface features where molecules might interact. These algorithms analyze geometric properties, volume, depth, and enclosure to locate candidate binding regions.

Step 4: Druggability Analysis

Not every binding site represents a viable therapeutic opportunity. Druggability analysis evaluates whether identified sites have appropriate characteristics for drug binding: suitable volume, appropriate hydrophobicity, hydrogen bonding potential, and accessibility. Sites scoring poorly on druggability metrics may be deprioritized.

Step 5: Molecular Design

With target sites identified and characterized, the design phase generates candidate molecules. Depending on the therapeutic strategy, this might involve:

• Small-molecule design or virtual screening
• Protein binder design (nanobodies, antibodies, mini-binders)
• Peptide design
• Other therapeutic modalities

Step 6: Docking and Interaction Analysis

Molecular docking predicts how candidate molecules might position themselves within binding sites. Interaction analysis examines the predicted contacts—hydrogen bonds, hydrophobic interactions, electrostatic forces—to understand which molecular features contribute to binding.

Step 7: Candidate Prioritization

The final computational step combines evidence from structure quality, binding predictions, interaction analysis, and other factors to rank candidates. This prioritization helps researchers decide which candidates warrant experimental validation.

Step 8: Experimental Validation

Prioritized candidates advance to laboratory testing. Experimental methods measure actual binding, assess biological activity, and generate the evidence needed to advance candidates through discovery. Computational predictions are hypotheses; experiments provide validation.