Supporting data for "A methodological approach to correlate tumor heterogeneity with drug distribution profile in mass spectrometry imaging (MSI) data"

Mass spectrometry imaging (MSI) has become a valuable tool in drug imaging because of its ability to provide a simultaneous spatial distribution of the drug and several other molecular ions present in the biological sample. An important application is the evaluation of homogeneity/heterogeneity of drug distribution in solid tumors. Solid tumors are known to be made up of different tissue subpopulations and their heterogeneity is supposed to have a direct and/or indirect influence on drug distribution. Hence, for further enhancement of penetration therapy performance, it is important to link a characterization of the tumor microenvironment with drug homogeneity. In this study, untargeted MSI data were used to understand the spatial heterogeneity within solid tumors, assessing its impact on the drug (paclitaxel) distribution. The proposed approach was applied to MSI datasets already analyzed, focusing on tumor drug distribution. Untargeted MSI datasets were collected on different tumor xenograft models (ovarian and colon cancer cell lines) pre-treated or not with anti-angiogenesis compound (bevacizumab). Our main data analysis steps involved: a) pre-processing of MSI data to make all biological samples directly comparable, b) unsupervised data clustering to find different tissue subtypes, c) quantification of drug heterogeneity using local indicators of spatial association (LISA) map and d) selection of important ion signals from identified clusters of interest using the spatial -aware statistical tools. Our clustering results show variation in tumor subpopulations and less spatial heterogeneity in the MSI data collected on samples treated with the anti-angiogenesis compound consistently with our previous data. The local spatial structures identified in drug ion LISA maps show a correlation with clusters identified using a clustering method. Using the right spatial method, we were able to reduce the number of false-positive ions selected and identified the one that shows relevant spatial patterns in different tissue subtypes. Finally, our overall study shows that there is a direct association in drug homogeneity and spatial arrangement of different tissue subtypes in a solid tumor.