Tumor antigens processed and presented by human leukocyte antigen (HLA) Class I alleles are important targets in tumor immunotherapies. Clinical trials showed that presence of CD8+ T cells specific to tumor associated antigens (TAAs) and tumor neoantigens is one of the main factors resulting in tumor regression. Affinity prediction of tumor antigen epitopes to HLA is an important reference index for peptide selection, which is highly individualized. In this study, we selected 6 CTAs (cancer-testis antigens) commonly used in cancer immunotherapy and top 95 hot mutations from the Cancer Genome Atlas for analyzing potential epitopes with high affinities to the common HLA class I molecules in white and East Asian population, respectively. The results showed that the overall difference in CTAs epitope prediction is small between the two populations. Meanwhile, there is a linear relationship between the CTAs peptide length and the relative overall epitope occurrence. However, the difference is bigger for epitopes prediction of missense mutations between the two populations. It is worth noting that, both in the two populations, the single point mutations with the highest incidences have the lowest epitope occurrences while the mutations with the highest epitope occurrences are with low mutation incidences. This may be the result of long-term selection by the host immunosurveillance. Frameshift/inframe indel mutation neoantigens are between CTAs and spot mutation neoantigens in the relationship between peptide length and predicted epitope number. Our results help provide clues for tumor antigen and epitope selection in cancer vaccine design.