Single-cell RNA sequencing (scRNA-seq) and single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq) provide complementary views of cellular states by capturing transcriptomic and chromatin accessibility landscapes, respectively [1]. Combining these modalities offers a comprehensive understanding of cellular functions and regulatory mechanisms. Here, we present multi-GAT, a model specifically designed for integrative analysis of scRNA-seq and scATAC-seq data using Canonical Correlation Analysis (CCA) followed by Graph Attention Network (GAT) to predict cell types. This approach leverages shared nearest neighbors and contrastive learning to enhance model performance. Multi-GAT effectively captures the complex relationships between transcriptomic and chromatin accessibility data, achieving robust cell type annotation across different single-cell modalities. The experimental results demonstrate that multi-GAT surpasses several baseline methods in accuracy, precision, and F1-score on the benchmark dataset.