How to Use Lacuna Cryptic Pocket Discovery

Lacuna is a structure-based cryptic-pocket discovery method for proteins whose relevant binding cavities may be absent from a single static conformation. It generates alternative conformations, detects cavities in the uploaded structure and sampled ensemble, and groups corresponding pockets across conformations so transient opening behavior can be measured directly.

On Neurosnap, researchers upload one PDB or mmCIF structure, select a sampling method, and choose how pockets are ranked and filtered. The resulting ranked pocket table reports pocket position, apo-to-open volume behavior, druggability, persistence, crypticity, and lining residues. Optional exports provide pocket visualizations, Boltz pocket constraints, and AutoDock Vina search boxes.

Lacuna first samples protein conformations with normal-mode analysis, implicit-solvent OpenMM molecular dynamics, or experimental Boltz-2 diffusion sampling. Normal-mode analysis is the recommended fast default. OpenMM provides force-field-based motion, while the upstream documentation cautions that Boltz currently generates conformers de novo from sequence and can produce structurally divergent, noisy ensembles.

A grid-based detector identifies cavities independently in the uploaded structure and every generated conformation. Lacuna then merges spatially corresponding pockets across the ensemble and computes volume range, peak open-state druggability, persistence, and crypticity. Crypticity increases when a pocket is absent or small in the uploaded structure but opens into a druggable cavity in sampled conformations.

The default learned strategy uses a fitted linear ranker over geometric and ensemble-derived features. The analytic alternatives emphasize crypticity, peak druggability, persistence, or a balance of druggability and persistence. These scores prioritize hypotheses within one protein; they are not calibrated for direct comparison between unrelated targets. Researchers should inspect the pocket geometry and lining residues alongside the ranking before using a site to guide docking, mutagenesis, or experimental follow-up.

• Multiple sampling methods: Normal-mode analysis, OpenMM, and Boltz-2 provide different speed and physical-model tradeoffs.
• Apo-aware pocket analysis: The uploaded structure is evaluated together with generated conformations, allowing transient opening behavior to be distinguished from constitutive pockets.
• Interpretation-ready ranking: Volume dynamics, druggability, persistence, crypticity, and lining residues are reported together for each retained site.
• Direct downstream artifacts: Optional pocket constraints, docking boxes, and pseudoatom structures support docking setup and structural review.