Application of data mining techniques on diabetes related proteins

Abstract

Genomic Data is growing very rapidly with the sequencing of genomes of various forms of life. To understand the overwhelming data and to obtain meaningful information, Data Mining techniques such as Principal Component Analysis and Discriminant Analysis are used for the purpose. Data Mining is basically used when the data is vast and there is need to extract the hidden knowledge in the form of useful patterns. The data set taken into consideration is protein data pertaining to diabetes mellitus obtained from a database. The task at hand was to find out in which species most of the diabetes related proteins exist. It so happened that most of these proteins were prevalent in Human Beings, House Mice and Norway Rat as they are all mammals and Human Beings have orthologs as House Mice and Norway Rat. Both these techniques prove that human beings show a variation from those of House Mice and Norway Rat which are similar in terms of the variation of protein attributes. This can also be inferred from statistical analysis by using histograms and bivariate plots. Other Data Mining Techniques such as Regression and Clustering can be used to further explore the above inference.

Appendix 1

The protein variates are:

• 1. Variate 1 is the length (L) of the protein in number of amino acids.

• 2. Variate 2 is the percent of basic amino acids in a given protein. The basic amino acids are H, K; R. percent basic is given by

  $$ \frac{{{\text{Number}}\,{\text{of}}\,{\text{basic}}\,{\text{amino}}\,{\text{acids}}}}{{{\text{Total}}\,{\text{number}}\,{\text{of}}\,{\text{amino}}\,{\text{acids}}}} \times 100 $$

• 3. Variate 3 is the percent of acidic/amide amino acids in a given protein. The acidic/amide amino acids are D, E, N, and Q. Percent acidic/amide is given by

  $$ \frac{{{\text{Number}}\,{\text{of}}\,{\text{acidic/amide}}\,{\text{amino}}\,{\text{acids}}}}{{{\text{Total}}\,{\text{number}}\,{\text{of}}\,{\text{amino}}\,{\text{acids}}}} \times 100 $$

• 4. Variate 4 is the percent of small and medium hydrophobic amino acids in a given protein. The small and medium hydrophobic amino acids are V, L, I, M. Percent hydrophobicity is given by

  $$ \frac{{{\text{Number}}\,{\text{of}}\,{\text{hydrophobic}}\,{\text{amino}}\,{\text{acids}}}}{{{\text{Total}}\,{\text{number}}\,{\text{of}}\,{\text{amino}}\,{\text{acids}}}} \times 100 $$

• 5. Variate 5 is the percent of aromatic amino acids in a given protein. The aromatic amino acids are F, Y, and W. Percent aromatic is given by

  $$ \frac{{{\text{Number}}\,{\text{of}}\,{\text{aromatic}}\,{\text{amino}}\,{\text{acids}}}}{{{\text{Total}}\,{\text{number}}\,{\text{of}}\,{\text{amino}}\,{\text{acids}}}} \times 100 $$

• 6. Variate 6 is the percent of small/polar amino acids in a given protein. The small/polar amino acids are A, G, S, T, P [32]. (Teresa K. Attwood et al. 2004). Percent small/polar is given by

  $$ \frac{{{\text{Number}}\,{\text{of}}\,{\text{small/polar}}\,{\text{amino}}\,{\text{acids}}}}{{{\text{Total}}\,{\text{number}}\,{\text{of}}\,{\text{amino}}\,{\text{acids}}}} \times 100 $$

• 7. Variate 7 is a measure of distance of a protein sequence from a fixed reference point.

  The distance is measured according to the formula:

  $$ {\hbox{Distance }}{\left( {\hbox{D}} \right)_{\rm{fixed}}} = { }\surd {\sum^{{2}0}}_{{\rm{i}} = {1}}{\left( {{{\hbox{O}}_{\rm{i}}} - { }{{\hbox{E}}_{\rm{i}}}} \right)^2} $$

  where O_i is the observed number of amino acid of type ‘i’ in the concerned protein and E_i, the expected number of amino acid of type ‘i’ in the same protein. E_i is L/20 considering all amino acid to be uniformly distributed in the protein. We refer to this point as the fixed reference point. Dfixed is square root of sum of squares from i = 1 to 20 of difference of observed and expected number of amino acids. Here it is considered fixed as Ei = L/20 is a constant for all the amino acids.

• 8. Variate 8 is the distance of a protein sequence from a variable reference point. The distance Dvar, globular has the same formula as that in variate 4 but the Ei is calculated according to the formula:

  $$ {\hbox{Ei }} = {\hbox{ fi}} {\hbox{x}} {\hbox{L}} $$

  where L is the length of the concerned protein in amino acids and fi is the average frequency of occurrence of the ith amino acid in the set of proteins that are of high sequence complexity [11]. Here this is considered variable reference point since fi changes for every amino acid and hence Ei changes.