Proteina-Complexa: Advanced Atomistic Binder Design Now Available on Vecura

What is Proteina-Complexa?

Proteina-Complexa is an advanced, fully atomistic generative model designed for de novo protein binder and scaffold engineering. Unlike traditional multi-stage workflows that decouple structure generation from sequence design, Proteina-Complexa uses a unified conditional flow-matching architecture to simultaneously sample backbone coordinates, side-chain orientations, and amino-acid sequences. By integrating inference-time search-based optimization strategies—such as best-of-N, beam search, and Monte Carlo Tree Search (MCTS)—the model significantly amplifies the generation of physically plausible, highly designable protein structures.

It helps users streamline the complex process of creating novel binding proteins from scratch, tailored to specific structural or ligand-based constraints. It is especially useful for researchers in drug discovery, enzyme design, and synthetic biology who need to design high-affinity binders against complex targets like proteins, small molecules, and carbohydrates.

What can users do with Proteina-Complexa on Vecura?

With Proteina-Complexa on Vecura, users can:

• Design de novo binders: Generate custom proteins targeting specific receptor surfaces, with optional hotspot conditioning to focus on desired epitopes.
• Engineer ligand-specific binders: Develop proteins capable of binding to small-molecule drugs or carbohydrate antigens, facilitating biosensor and drug-conjugate development.
• Perform motif scaffolding: Incorporate catalytic or functional binding motifs into stable de novo folds, with or without ligand context, for precision protein engineering.
• Optimize designs in real-time: Leverage test-time compute algorithms to filter and rank candidates based on confidence metrics, hydrogen-bond networks, and predicted structural stability.

What the output means

The output provides a comprehensive candidate pool consisting of PDB structure files, corresponding optimized amino-acid sequences, and a detailed metrics table. These metrics include interface predicted aligned error (ipAE), interface RMSD, and designability scores, which help researchers identify the most promising candidates.

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

The ability to computationally design high-affinity protein binders has historically been limited by the computational cost and the disjointed nature of structural and sequence prediction pipelines. Proteina-Complexa addresses these bottlenecks by moving toward an unified, atomistic generative approach that is reinforced by heavy inference-time compute. This shift allows for unprecedented hit rates in in-silico design—demonstrated by successes in kinase mini-protein and carbohydrate binder targets—thereby accelerating the timeline from concept to experimental testing.

By enabling this sophisticated model through a guided workflow on Vecura, researchers can bypass the complex technical infrastructure and GPU orchestration typically required for such intensive generative tasks, allowing them to focus entirely on scientific discovery.

• Developed by: NVIDIA
• Source: Official GitHub repository
• Reference: Scaling Atomistic Protein Binder Design with Generative Pretraining and Test-Time Compute (ICLR 2026)