Our Research Projects

Evaluating Neural Encoder-Decoder Models for Period-Avoiding DNA Storage Sequences

This is a project which is currently making use of HPC facilities at Newcastle University. It is active.

Project Contacts

For further information about this project, please contact:


Project Description

This project investigates the use of deep neural networks (autoencoders) for DNA data storage encoding. The model learns to encode binary information into DNA sequences that satisfy biological constraints (period-avoidance constraints) while maintaining high reconstruction accuracy. We train convolutional neural network-based encoder-decoder architectures with differentiable constraint penalties, requiring GPU acceleration for longer sequence lengths and extended training.


Software or Compute Methods

Python 3.10+, PyTorch (deep learning framework), running on GPU nodes. The workflow involves training CNN-based autoencoder models with approximately 30,000-50,000 training epochs per experiment. We require GPU acceleration (CUDA) for matrix operations on sequence data. No specialised licensed software is needed all tools are open-source.