Effect of mass segment size on polymer ToF-SIMS multivariate analysis using a universal data matrix

This work is part of a comprehensive research program to understand the informatics issues associated with high resolution surface analysis methods that are becoming essential for understanding how materials interact with biology. We have shown that advanced informatics methods can extract more information from surface analysis experiments than the traditional linear PCA methods commonly employed. These advanced methods can reliably separate polymers and other materials with very similar chemical structures which traditional methods cannot achieve. Building on our prior work, we report the effects of finer and coarser binning (1 m/z, 0.1 m/z and 0.005 m/z) of ToF-SIMS data on the ability of informatics methods to discriminate between chemically similar polyamide polymers. We show that the linear multivariate analysis methods PCA, HCA and MCR fail to discriminate mass discretised matrix data due to high levels of variance. In contrast, self-organising maps (SOMs), optimised for mass segment size, prove very tolerant to variance and noise and exhibit excellent classification efficiency for the chemically similar polyamide groups. Our results provide an important step in the development of a new paradigm in which analysis of data from ToF-SIMS and other analytical methods, such as Raman/SERS, can be conducted in a fully automated fashion.