We are developing a comprehensive language foundation model tailored to unravel the intricacies of mRNA diversity, encompassing alternative Polyadenylation (APA) and alternative splicing (AS), within single-cell RNA sequencing data. Our goal is to harness extensive single-cell and spatial transcriptome datasets to decipher APA/AS patterns and gene expression simultaneously, expanding the scope beyond conventional single-cell expression analysis. Through fine-tuning tasks, the model will be able to execute a series of downstream prediction tasks, offering a multifaceted perspective on gene regulation at single-cell resolution and within spatial contexts. This work will have the potential to advance RNA-based therapeutic strategies targeting diseases by directing efforts towards modulating APA and AS.