Transformer-Based Operon Prediction Using Textual Representations of Gene Pairs Supplementary Dataset

Transformer-Based Operon Prediction Using Textual Representations of Gene Pairs This repository accompanies the manuscript "Transformer-Based Operon Prediction Using Textual Representations of Gene Pairs". Overview We present a transformer-based framework for operon prediction that recasts gene-pair classification as a natural language processing task. By serializing genomic features into structured textual descriptions, our approach enables pre-trained language models to classify whether two adjacent genes belong to the same operon. Directory Structure Data The folder includes CSV files containing gene-pair examples used for training and evaluation. Each row corresponds to a gene pair with serialized textual input and operon membership label. The structure is: | genome_id | gene1 | gene2 | text | label ||-----------|---------------|---------------|------------------------------------------------------------------------------------------------|-------|| 169963 | 169963.11.peg.2 | 169963.11.peg.3 | "Gene A is 1146 bp long... Gene B is 1344 bp long..." | 0 | - `text`: the serialized paragraph describing the gene pair (as detailed in the paper’s Methods).- `label`: binary indicator (1 = same operon, 0 = not same operon). > Note: Gene features were derived using the BV-BRC API and annotations. See the manuscript for full details. Reproducibility The experimental setup and evaluation regimes are described in detail in the manuscript. While we do not include the full training notebook here, all necessary data and preprocessing details are provided to support reproducibility. Citation If you use this code or dataset, please cite the paper.