<p>This paper considers variable selection and parameter estimation in high-dimensional heteroscedastic regression models using a Bayesian approach. Utilizing the Spike-and-Slab LASSO prior, which is suitable for high-dimensional data, we establish the large-sample posterior inference properties under a non-symmetric squared loss function. To demonstrate the efficacy of the proposed method in dealing with high-dimensional heteroscedastic data, we examine scenarios of error homogeneity and two additional heteroscedastic structures. Our method exhibits superior performance in variable selection and parameter estimation, and can effectively identify heteroscedasticity at various non-symmetric levels. Finally, we apply our method to the gene (eQTL) dataset, obtaining smaller prediction errors compared to existing methods.</p>