DNA-origami-stability-prediction

Predicting DNA origami stability in physiological media by machine learning

Judith Zubia-Aranburu¹, Andrea Gardin¹, Lars Paffen, Matteo Tollemeto, Ane Alberdi, Maite Termenon, Francesca Grisoni^*, Tania Patiño Padial^*
1These authors contributed equally to this work.
^*Corresponding authors: j.c.m.v.hest@tue.nl, f.grisoni@tue.nl, l.brunsveld@tue.nl.

Disclaimer

The work has been submitted for peer-review and it might not represent the final version, the code and/or the content of the paper can be subjected to changes.

Abstract

DNA origami nanostructures offer substantial potential as programmable, biocompatible platforms for drug delivery and diagnostics. However, their structural stability under physiological conditions remains a major barrier to practical applications. Stability assessment of DNA origami nanostructures has traditionally relied on image-based and empirical approaches, which are time-consuming and difficult to generalize across conditions. To address these limitations, we developed a machine learning approach for DNA origami stability prediction, based on measurable physicochemical parameters. Using dynamic light scattering (DLS) to quantify diffusion coefficients as a proxy for structural integrity, we characterized over 1400 DNA origami samples under varying physiologically relevant variables: temperature, incubation time, MgCl2 concentration, pH, and DNase I concentrations. The predictive performance of the model was confirmed using an independent set of samples under previously untested conditions. This data-driven approach offers a scalable and generalizable framework to guide the design of robust DNA nanostructures for biomedical applications.

Content

This repository contains the code used to apply the machine learning pipeline described in the main paper.

This repository is structured in the following way:

• /datasets/ : folder containing the datasets used to train/test/validate the models in our experiments.
• /experiments/ : folder containing the output data collected from the experiments described in the paper.
• /figures/ : folder containig high resolution figure as reported in the main paper.
• /script/ : folder containig the script for running a replicate or a new experiment using the described ML set up.
• /src/origamiregressor/ : main folder containig the code modules defining the package.
• environment.yaml : the environment file to create and install the package.
• pyproject.toml : the setup file for installing the package.

Intallation

The package and all the needed dependencies can be installed with the provided env.yaml file. The installation was tested on Ubuntu 22.04.3.

conda env create -f environment.yaml

Running Experiments

A quick tutorial on how to run the experiments, to reproduce and/or test the results, is given in the ./experiment/ folder.

How to cite

You can currently cite our work from our preprint:

Predicting DNA origami stability in physiological media by machine learning.
Judith Zubia-Aranburu, Andrea Gardin, Lars Paffen, Matteo Tollemeto, Ane Alberdi, Maite Termenon, Francesca Grisoni, Tania Patiño Padial bioRxiv, 18 July, 2025. DOI: https://doi.org/10.1101/2025.07.18.665506