This dissertation investigates marker gene discovery and authentication using Explainable AI (XAI) in transcriptomics analysis. While machine learning (ML) and deep learning (DL) methods achieve high accuracy in cell-type classification, they often lack interpretability. XAI enhances transparency, enabling the identification of key genes driving classification. We introduce XAI-driven pipelines for RNA-seq and Spatial Transcriptomics (ST). For RNA-seq, we develop CNN-CAM, which combines Convolutional Neural Networks (CNNs) and Class Activation Maps (CAMs) to extract marker genes. For ST, we propose GSI, an autoencoder-based pipeline for cell-type clustering in multi-sample ST datasets. Given the lack of robust ST classification methods, we introduce an ML-based authentication pipeline to validate clustering and extract reliable marker genes. Our findings show that XAI enhances ML/DL models, enabling biologically meaningful marker gene identification. This work advances transcriptomics analysis, providing a scalable framework for biomarker discovery, disease research, and precision medicine.