About Model

Purpose

• This model was developed to practice development of ncRNA sequence generation
• It is highly experimental, and should not be used in any real-world environments where accuracy is paramount

Methodology

• We utilize GNN architecture to predict RNA sequences which would fold towards a given structure
• The idea is that a researcher might want to have a ncRNA with a specific structure that will fold into a desired shape
• They might know what structure would make this shape, but they do not know the sequence that makes that structure

Training Data

• Trained on RFAM seed alignments RF00001–RF00500 (~500 RNA families, Stockholm format)
• Data sourced from the EBI RFAM REST API; see COLAB_SETUP.md for download instructions

Training Details

• Teacher forcing ratio is linearly annealed from 1.0 (epoch 1) to 0.1 (final epoch) to reduce exposure bias
• Early stopping with patience of 15 epochs prevents overfitting
• Train/validation split is a seeded random 80/20 split across all families

Usage

• See COLAB_SETUP.md for step-by-step instructions to train and run inference on Google Colab

AI Disclosure

• The code in this repo was written with assistance from Claude Code CLI tool.
• Human oversight was present - but I am not a biology subject matter expert.
• Model inferences should be used with caution, as there is risk of non-scientfic biases causing error.