This corroborates selleck chemicals the findings presented elsewhere [17]. For example, from the S. acidocaldarius analysis, we found that some up-regulated proteins contained specific codon usage pattern of aspartic acid. The specific codon usage pattern of aspartic acid could thus promote up-regulation of proteins; this is a hypothesis that is testable in the laboratory with important bioengineering applications. In contrast, the specific codon usage pattern of proline may reduce the translational efficiency, since proline is avoided in some up-regulated proteins. We also found that aspartic acid and proline were all correlated with the up-regulation of gene function L (Replication; recombination and repair). One possible explanation of this finding is that the up-regulation of those genes combined the two effects just described. Indeed, when evaluating the gene function distribution of the up-regulated proteins being rich in aspartic acid and poor in proline at the same time, only proteins belonging to gene function L (replication; recombination and repair) are conserved. The proposed bioinformatic approach may shed light on further investigations of the mechanisms of protein regulation. In previous studies, e.g., [17] and [18], synonymous codons were randomly substituted, and the substitutions resulted in significant protein regulation. However, those studies were not able to show which specific codons play a key role in protein regulation. This issue can be tackled by the proposed approach. By investigating the codon usage pattern of those amino acids which are tightly correlated with protein regulation, the effect of codon usage on protein regulation can be more precisely evaluated, which may lead to a better understanding of the mechanisms of protein regulation and applications in gene optimization. Furthermore, protein regulation may also be affected by amino acid properties, since we identified an amino acid through amino acid composition comparison. However, as only one amino acid was identified in this part of our analysis, it was difficult to further associate amino acid properties with protein regulation. One explanation for this may be that the effect of amino acids properties on the protein regulation is likely to be less significant compared to codon usage pattern correlated translational efficiency changes. Another reason may be that amino acid preference was obscured by the noise contained in the proteomic data. To reduce the affect of interference and noise, we have used relatively high threshold (p < 0.01) to identify the significant differences in the current analysis. By increasing the p-value threshold, one may identify more amino acids, but without significantly affecting our conclusions here. For example, alanine can be identified at the p < 0.05 level from the S. solfataricus proteomic dataset.