Towards batch correction for GC-IMS data

Abstract

Gas Chromatography Ion Mobility Spectrometry (GC-IMS) is a fast, non-expensive analytical technique that allows obtaining relevant chemical information from vapor mixtures. However, the technique presents some difficulties that should be solved to ensure reliable and reproducible results, namely: 1) data exhibits simultaneously high dimensionality and sparsity on their chemical information content, 2) data samples must usually be corrected even within a batch because of baseline and misalignment problems, 3) additional data corrections must be performed to prevent from chemical fingerprinting variations among batches. In this work, we have acquired data from two different batches (A and B) of ketone mixtures (2-Butanone, 2-Pentanone, 2-Hexanone, and 2-Heptanone). The analytical method for batch A and B was the same, except for the value of carrier gas flow parameter, which was approximately doubled for batch B. We have addressed problems 1) and 2) independently for each batch, obtaining as a result two peak tables. 3). Common peaks present in batches A and B were found after scaling the retention time axis of batch B and perform k-medoids clustering. Using this information, test data from batch B has been corrected through a linear transformation.