FMQA for RNA Inverse Folding

This repository provides the implementation of the Factorization Machine with Quadratic-Optimization Annealing (FMQA) applied to the RNA inverse folding problem.

Features

• Binary-integer encoding (one-hot, domain-wall, binary, unary)
• Nucleotide-to-integer assignment
• Ensemble defect-based objective function (ViennaRNA)
• FMQA optimization

Requirements

• Python 3.12
• ViennaRNA 2.7.2
• amplify >= 1.4.1
• torch >= 2.0

Setup

Set your Amplify token as an environment variable.

Mac / Linux

export AMPLIFY_TOKEN=your_token_here

Usage

Run the script with the required arguments:

python fmqa_rna.py --encoding one-hot --base_allocation AUGC

Output

The script prints the optimization progress:

FMQA cycle 0: found y = ...
FMQA cycle 1: found y = ...
...
=== Finished ===
encoding        : one-hot
base_allocation : AUGC
target_structure: ..((((((((.....)).))))))..
best_value      : 0.123456

Citation

If you use this package in your work, please cite the following papers (Shuta Kikuchi, and Shu Tanaka, 10.48550/arXiv.2602.16643)

Shuta Kikuchi, and Shu Tanaka, "Factorization Machine with Quadratic-Optimization Annealing for RNA Inverse Folding and Evaluation of Binary-Integer Encoding and Nucleotide Assignment," arXiv, arXiv:2602.16643, 2026, doi: 10.48550/arXiv.2602.16643.

@article{kikuchi2026factorization,
  title={Factorization Machine with Quadratic-Optimization Annealing for RNA Inverse Folding and Evaluation of Binary-Integer Encoding and Nucleotide Assignment},
  author={Kikuchi, Shuta and Tanaka, Shu},
  journal={arXiv preprint arXiv:2602.16643},
  year={2026},
  doi={10.48550/arXiv.2602.16643}
}