Figure 3

Dendrograms from hierarchical cluster analysis of spectra from CMLS-F4-QC data set labelled by batch, processed before (left) and after (right) Partek Batch Removal. QC spectra peak intensities from different points in the analytical process were used to generate dendrograms from hierarchical cluster analysis (Spearman rank dissimilarity metric with average linkage). Each spectrum was labeled for the batch in which it was processed (A through K). The dendrogram on the left is from analysis of spectra before Partek Batch Remover was applied to the data set. Spectra cluster in nodes according to the batch in which they were processed. Two large clusters are evident. One with spectra from batches A through F. The second covering batches G through K. An unanticipated instrument adjustment had to be made between the sixth and seventh batch, which is noticed in the analysis. The dendrogram on the right shows the hierarchical cluster analysis of the same data after Partek Batch Remover was used to reduce the contribution of batch effect technical variance. Spectra no longer cluster by batch in which they were processed, and spectra from before and after the instrument maintenance are inter-mingled across the 2 major nodes.