In this study, 414 whole protein-coding sequences (238 004 codons) of alternatively spliced genes of human chromosome 1 have been employed to explore the patterns of codon usage bias among genes. Overall codon usage data analysis indicates that G- and C-ending codons are predominant in the genes. The base usage in all three codon positions suggests a selection-mutation balance. Multivariate statistical analysis reveals that the codon usage variation has a strong positive correlation with the expressivities of the genes (r=0.5790, P＜0.0001). All 27 codons identified as optimal are G- and C-ending codons.Correlation analysis shows a strong negative correlation between the gene length and codon adaptation index value (r=-0.2252, P＜0.0001), and a significantly positive correlation between the gene length and Nc values (r=0.1876, P＜0.0001). These results suggest that the comparatively shorter genes in the genes have higher codon usage bias to maximize translational efficiency, and selection may also contribute to the reduction of highly expressed proteins.