QUBODock: A Pip-Installable QUBO Tool for Ligand Pose Generation

TL;DR

QUBODock is a Python tool that formulates ligand pose generation as a QUBO problem, enabling efficient GPU-accelerated solutions and flexible integration with external scoring functions for drug discovery.

Contribution

It introduces a pip-installable, GPU-accelerated QUBO-based ligand pose generation tool with a minimal interface and open-source release for benchmarking and development.

Findings

Efficient pose generation on CPU and GPU.

Flexible pairing with external scoring functions.

Open-source implementation for benchmarking.

Abstract

We present QUBODock, a pip-installable tool that formulates ligand pose generation as a Quadratic Unconstrained Binary Optimization (QUBO) problem and solves it efficiently on CPU or GPU. QUBODock focuses exclusively on pose generation and deliberately excludes any built-in scoring function, allowing researchers to pair its poses with external scorers of their choice. The software provides a minimal, reproducible interface for (i) protein-ligand structure ingestion and preprocessing, (ii) QUBO model construction from geometric/compatibility constraints, and (iii) decoding solutions into candidate poses for downstream ranking. Implemented in Python with GPU acceleration, QUBODock emphasizes usability and reproducibility: it is distributed on PyPI and can be installed with a single command. We release the source to support benchmarking, teaching, and method development around QUBO-based docking pose generation.