Detection and quantification of embedded minor analytes in three-way multicomponent profiles by evolving projections and internal rank annihilation

Modern analytical instrumentation often leads to data arrays with more than two dimensions or directions. Such N-way data (N > 2) needs special resolution methods for optimising the amount of analytical information. In this work, a new interactive method designed to work with three-way data is presented. The method, evolving projections by optimised search (EPOS), presents a combined graphical and numerical way of resolving a three-way data array into the analytical profiles of the pure analytes. The method involves an internal rank annihilation step, which can be performed in several ways. The graphic interactive procedure used in this work compares favourably with Lorberʹs noniterative rank annihilation method. Thus, our method is significantly better for resolution of analytes with low relative concentration, especially in the presence of heteroscedastic noise. The EPOS method is tested on several simulated data sets to assess its performance. A peak purity example is carried out to show a case where two-way methods are unable to provide a unique solution, whereas EPOS gives correct results.