PanGen-AI: An Integrated Deep Learning and Multi-Track Genome Visualization Framework for Pangenomic Data Analysis

The transition from linear reference genomes to graph-based pangenomes, coupled with the rise of artificial intelligence, requires modern biologists to understand highly complex computational structures. However, a significant gap exists between wet-lab biological training and the algorithmic foundations of modern bioinformatics. Here, we present PanGen-AI , a comprehensive, open-source 8-module Python framework designed to bridge this gap. PanGen-AI integrates diverse computational engines, de Bruijn graph sequence assembly, a PyTorch-based 1D convolutional neural network (CNN) for variant impact prediction, Burrows-Wheeler Transform indexing, Needleman-Wunsch alignment, CRISPR-Cas9 design, and 3D protein structure visualization. The framework culminates in an interactive multi-track genome browser that dynamically overlays AI-derived variant saliency maps, CRISPR target sites, and classical gene annotations onto a unified genomic coordinate system. Deployed via a lightweight Streamlit interface, PanGen-AI serves as both a scalable prototyping environment for automated genomic workflows and a translational tool to demystify computational models in mechanobiology and immunometabolism.