Accelerate Antibody Discovery with AntiDIF on Vecura

What is AntiDIF?

AntiDIF (Antibody-specific Discrete Diffusion for Inverse Folding) is a specialized discrete-diffusion model designed to generate novel antibody amino-acid sequences that fold into a specific, prescribed backbone geometry. By fine-tuning the RL-DIF architecture on antibody-specific structural data, the model effectively addresses the challenge of creating diverse antibody candidates while maintaining high sequence-recovery rates. It is particularly valuable for researchers aiming to explore a wider design space during antibody discovery campaigns, moving beyond the limitations of general-purpose inverse folders.

What can users do with AntiDIF on Vecura?

With AntiDIF on Vecura, users can:

• Generate novel antibody variable-domain sequences directly from a provided heavy/light chain PDB backbone.
• Improve the diversity of sampled sequences compared to traditional autoregressive methods.
• Perform targeted redesign by restricting sequence generation to specific CDR loops or hotspot residues using custom free_positions.
• Easily manage the design process through a guided interface without needing to configure complex, underlying infrastructure.

What the output means

The output provides a comprehensive set of predicted sequences, detailed sequence recovery metrics, and a diversity score that quantifies how varied the generated candidates are for a given backbone.

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

Why this matters

The traditional antibody design process often faces a bottleneck where computational tools produce sequences with limited structural or sequence variation, restricting the scope of experimental screening. By enhancing inter-sample diversity while strictly adhering to a defined backbone, AntiDIF allows researchers to explore a broader, more promising candidate space.

This capability empowers scientists to generate more robust, diverse panels of antibody candidates, significantly accelerating the initial stages of therapeutic antibody discovery.

• Developed by: Nikhil Branson (OXPIG)
• Source: Official GitHub Repo
• Reference: AntiDIF Preprint (bioRxiv)