An accurate clustering algorithm for fast protein-profiling using SCICA on MALDI-TOF

In this paper we propose an accurate clustering algorithm as the necessary step of the Single Channel Independent Component Analysis (SCICA) in the context of the fast extraction of protein profiles from the mass spectra (MALDI-TOF) data. In general K-means clustering is employed for clustering of the basis vectors. However given its iterative and statistical nature, convergence to the same clusters for the same data sets is not always guaranteed making it inaccurate, especially in protein-profiling where reliability of the bio-marker based disease detection and diagnosis depend immensely on the reliability of the clustering algorithm. Furthermore the proposed algorithm does not involve any arithmetic computations helping expedite the entire SCICA process.