Spliceformer v1.0.0 - Transformer Model for Splice Site Prediction

This initial release of Spliceformer provides an end-to-end framework to train and perform inference with a 45k-context transformer model designed for splice site prediction and genetic variant analysis. The repository includes code for dataset construction, model training, fine-tuning on RNA-Seq annotations, and analysis of splice site prediction scores. Pre-trained models are also available for quick start predictions. Accompanying Jupyter notebooks guide users through reproducible workflows, including constructing datasets, training, fine-tuning, and analyses, such as spliceAI vs. transformer score comparison and UMAP visualization.