An algorithmic approach to understand trace elemental homeostasis in serum samples of Parkinson disease.

A classical problem in neurological disorders is to understand the progression of disorder and define the trace elements (metals) which play a role in deviating a sample from normal to an abnormal state, which implies the need to create a reference knowledge base (KB) employing the control samples drawn from normal/healthy set in the context of the said neurological disorder, and in sequel to analytically understand the deviations in the cases of disorders/abnormalities/unhealthy samples. Hence building up a computational model involves mining the healthy control samples to create a suitable reference KB and designing an algorithm for estimating the deviation in case of unhealthy samples. This leads to realizing an algorithmic cognition-recognition model, where the cognition stage establishes a reference model of a normal/healthy class and the recognition stage involves discriminating whether a given test sample belongs to a normal class or not. Further if the sample belongs to a specified reference base (normal) then the requirement is to understand how strong the affiliation is, and if otherwise (abnormal) how far away the sample is from the said reference base. In this paper, an exploratory data analysis based model is proposed to carry out such estimation analysis by designing distribution and parametric models for the reference base. Further, the knowledge of the reference base in case of the distribution model is expressed in terms of zones with each zone carrying a weightage factor. Different distance measures are utilized for the subsequent affiliation analysis (City block with distribution model and Doyle's with Parametric model). Results of an experimental study based on the database of trace elemental analysis in human serum samples from control and Parkinson's neurological disorder are presented to corroborate the performance of the computational algorithm.