# NanoDeNovo: De Novo Design of Anti-Poliovirus I Sabin Strain Nanobodies by Semi-Automated Computational Pipeline

## TL;DR

This paper introduces a computational pipeline to design nanobodies targeting poliovirus, offering a new approach for developing antiviral therapies.

## Contribution

A semi-automated computational pipeline for de novo design of anti-poliovirus nanobodies is developed and validated.

## Key findings

- Three humanized nanobodies showed strong binding to VP3 with binding free energies ranging from −37.66 to −48.62 kcal/mol.
- Humanized nanobodies exhibited superior stability and structural similarity to experimentally validated designs.
- The pipeline produced nanobodies with favorable physicochemical properties and high solubility.

## Abstract

Despite global vaccination efforts, poliomyelitis continues to cause paralytic cases, highlighting the need for alternative therapeutic approaches. Nanobodies offer significant advantages over conventional antibodies due to their small size, stability, and low immunogenicity, yet few have been developed specifically against poliovirus. This study presents a fully computational pipeline for de novo design of nanobodies targeting Virus Protein 3 (VP3) of the Poliovirus I Sabin strain. Our integrated approach employed Ig-VAE for scaffold generation, ProteinMPNN and RFantibody for sequence design, tFold-Ab/Ag for structure prediction, multi-platform molecular docking (Rfantibody, Rosetta3, ClusPro2, ReplicaDock 2.0), molecular dynamics simulations, and humanization tools. The pipeline identified three humanized nanobodies (scFv-0389-304-6H, scFv-0389-459-5H, and scFv-0743-166-7/H) that demonstrated strong binding to VP3 with binding free energies of −37.66 ± 10.35, −40.11 ± 20.01, and −48.62 ± 11.21 kcal/mol, respectively. All designs exhibited favorable physicochemical properties and high solubility. Notably, nanobodies humanized prior to CDR-loop design (scFv-0743-166-7/H) showed superior stability, binding affinity, and structural similarity to experimentally validated nanobodies. This work demonstrates the feasibility of a fully computational approach for designing promising nanobodies against viral targets, providing an alternative to traditional methods with potential applications in drug design.