Weighted Manhattan distance classifier; SELDI data for Alzheimer’s disease diagnosis
Mass Spectrometry (Surface Enhanced Laser Desorption Time of Flight (SELDI-TOF) assay technique) for proteomics is based on the consistency and reproducibility of protein/peptide expressions. In this study, we opine that mining collections of mass spectra data instead of detailed study of individual ions generated in the course of Mass Spectrometer assay process, will generate discriminative factors for the diagnosis of Alzheimer's Disease (and other diseases in general). This model; Weighted Manhattan Distance Classifier (WMDC), classifies a test vector to the stage label of the most significant train vector to it using Manhattan Distance function and thereafter, classifies a test data point (a collection of test vectors) to the disease stage having the majority of most significant train vectors in it. The disease severity is categorized as normal/control, mild and acute impaired stages, each of which contained 20 SELDI-TOF analysis results. In all, the database contained 60 assay results of saliva analytes or protein source samples under 3 proteinChips; CM10, IMAC30 and Q10. Each laboratory experiment was performed with either low (1800 nJ) or high (4000 nJ) laser energy bombardment level. 90% classification result was obtained with a probability of 0.075 for committing type II error (that is, a test power of 0.925).
Anyaiwe, O. E. D., Singh, G. B., Wilson, G. D., & Geddes, T. J. (2017). Weighted manhattan distance classifier; SELDI data for alzheimer's disease diagnosis. Paper presented at the 257-262. doi:10.1109/CEC.2017.7969321
2017 IEEE Congress on Evolutionary Computation (CEC), San Sebastián, Spain, June 5-8, 2017.