Proteina-Complexa

Proteina-Complexa designs protein binders against a target structure with NVIDIA BioNeMo's generative pipeline.

Inputs

Upload one protein target structure in PDB format. The target residue field follows Proteina-Complexa notation:

• A uses all residues from chain A
• A1-150 uses residues 1 through 150 from chain A
• A1-100,B1-50 uses ranges from multiple chains

Hotspot residues are optional. When provided, they should be comma- or space-separated residue identifiers such as A35,A57,A91.

Pipeline Modes

Full design runs the source binder workflow: generation, filtering, evaluation, and analysis. It returns generated structures, reward tables, evaluation tables, logs, and source files produced by the run.

Generate and filter runs the generation and filtering stages only. This is useful when you need the initial generated candidates and reward table without the heavier evaluation stages.

Key Settings

• Diffusion steps defaults to 400, matching the source binder design configuration.
• Best-of-N is the default search algorithm, with 2 replicas.
• The default generation batch size is 16.
• The default random seed is 5.
• The default evaluation sequence redesign count is 2.

Outputs

ProteinIQ returns the generated PDB structures, CSV result tables, run logs, and downloadable source files produced during the job. The top-sample and reward tables are shown as spreadsheets when available.

Notes

This initial ProteinIQ tool focuses on protein-target binder design. Ligand binder design and antigen-motif epitope scaffolding are tracked separately because they require different source configurations and model assets.