MULTIFRACTAL ANALYSIS OF PROTEIN AGGREGATES TO DERIVE PROTEIN-SPECIFIC SIGNATURE

Deriving a property of a protein that is unique to it has well known significance since the study on ab initio model based derivation of protein structure where uniqueness of protein sequence is taken as the source of specificity of protein structure. In this direction, Heat denatured protein aggregates (HDPA) of proteins were studied with an objective to derive some multi-fractal markers specific to constituent protein that may be further utilized to extract information of the seed protein. Since Ordinary microscopic images of aggregates were analyzed to extract Intensity Level-based Multifractal Dimension (ILMFD) features. ILMFD features include four different features, perimeter fractal dimension (ILMFDP), perimeter-area relationship (ILMFDPAR), Area fractal dimension (ILMFDA) and Perimeter-area fractal dimension (ILMFDPA) that were calculated using fractal computations considering perimeter, and area of aggregate images. Feed forward backpropagation network was used to classify the proteins using different ILMFD parameters. It was found that ILMFD features could discriminate the proteins used in our study, that points to their potential to serve as unique property or marker of a protein. Further to validate the results, the outputs from ANN were clustered, and the outputs clustered in the largest cluster were found to significantly improve the result in class decision given by ANN.