The accurate detection of outliers in gene expression datasets plays a crucial role in the unraveling of intricate biological processes. This research introduces "SymNOM-GED," an innovative algorithm for outlier mining in gene expression datasets, with a focus on Esophageal Squamous Cell Carcinoma (ESCC). SymNOM-GED leverages symmetric neighbor to effectively identify outliers by considering local and global gene expression patterns. Extensive experiments demonstrate that SymNOM-GED outperforms existing algorithms in terms of accuracy, robustness, and scalability. The algorithm's performance is validated using clustering coefficient, graph density, and modularity, confirming its superiority. SymNOM-GED's precise and reliable outlier detection capabilities contribute significantly to bioinformatics research, offering insights into gene expression patterns in diverse biological contexts.