Proteins constitute about 50% of the dry weight of most cells and are the most structurally complex macromolecules known. Proteins can be classified in different manner but for the purpose of this study we classified them as membrane (part of either cellular or organelle membrane; MPs) and non-membrane (located outside the membrane; nMPs) proteins. Amino acids are the building block of a protein and their composition determines the overall properties and stability of a protein. Many previous studies have shown how amino acid composition can be successfully applied to protein sequence analysis, including prediction of structural class (Zhang et al., 1992), discrimination of intra- and extra cellular proteins (Nakashima et al., 1994), prediction of sub-cellular location (Cedano et al., 1997). It was suggested that composition differences are a consequence of different requirements for protein folding, stability and transportation.

The recent increase in the number of whole genome sequences has made the analysis of the corresponding proteomes possible. So far the amino acid composition of both the prokaryotic and eukaryotic proteomic databases have been explored separately for different purposes such as determination of sequence length (Gerstein, 1998a), identification of conserved sequences (Sobolevsky et al., 2005); elucidation of simple sequences (Subramanyam et al., 2006) etc. However, till now the comparative analysis of their non-membrane proteins (nMPs) have not been carried out to determine the overall amino acid compositional differences. This computational study is performed to develop the amino acid distribution of proteins as a tool to identify the proteins frequently undergo mutations and largely responsible for the pathogenicity of the organism.

The dataset was curated manually from the sequences extracted from PSORT (Rey et al., 2005), eSLDB (Pierleoni et al., 2007) and RefSeq (Pruittet et al., 2005) databases. Only the experimentally annotated entries were extracted from PSORT database. From the RefSeq database, we used microbial (microbial1.protein.faa.gz; 05/11/2009) and eukaryotic (vertebrate_mammalian1.protein. faa.gz; 05/11/2009 & vertebrate_other1.protein.faa.gz; 05/10/2009) sequence release files for construction of the experimental dataset. Protein sequences flagged as putat ive, hypothet ical , potent ial , uncharacterized, similar to the predicted protein, membrane, porin, receptor are deleted from the initially downloaded RefSeq sequence release files in the preparation of experimental dataset. The prokaryotic sequence dataset was created by merging the sequence entries from PSORT db and refseq dataset after appropriate deletions. Similarly, the eukaryotic dataset was prepared after deleting and merging the sequence entries from eSLDB and refseq dataset.

The entire dataset used for computing the composition of 20 amino acid residues comprised of prokaryotic (63644) and eukaryotic (88400) nMP sequences. The amino acid composition for the prepared datasets was computed using the number of amino acids of each type and the total number of residues. It is defined as Residue composition (%) (r) = Σn_r/N X100 (1) where 'r' stands for any one of the 20 amino acid residue. Σn_r is the total number of residue of each type and N is the total number of residues in the dataset.

The amino acid compositional distribution between prokaryotic and eukaryotic nMPs was computed using eq. (1). The prokaryotic nMPs shows the dominant occurrence of a non-polar amino acid 'Ala' (σ = 0.45) while the eukaryotic nMPs predominantly possess the polar amino acids 'Lys' (σ = 0.66), 'Ser' (σ = 0.60) and 'Cys' (σ = 0.29) (Figure 1). In prokaryotic nMPs, the high frequency of short side-chained non-polar aliphatic amino acid 'Ala' may be due to various possibilities such as its over-representation in highly expressed proteins (Tats et al., 2006), its role in determining the cleavage of N-terminal formyl methionine (Solbiati et al., 1999), its role in assisting the entrance of the nascent peptide chain into the ribosomal tunnel (Tenson et al., 2002) and in helix-helix packing (Eyre et al., 2004). Though 'Ala' might perform the similar functions in both prokaryotic and eukaryotic nMPs but its higher frequency in nMPs probably related to the higher proportion of prokaryotic helical nMPs.

The eukaryotes show the high occurrence of positively charged polar residue 'Lys' in their nMPs repertoire. This positively charged residue helps in the secretion of proteins through the membrane via interaction with export machinery and signal recognition particles (vonHeijne, 1984). The overabundance of 'Ser' in eukaryotic nMPs may be due to their ability to form H-bonds and stabilizing the helices (Subramaniam et al., 2006). In particular, the two-fold higher 'Cys' of eukaryotic nMPs compared to prokaryotic nMPs most probably compensates for their lower hydrophobicity (D'Onofrio et al., 1999).