Figure 2.

Overview of IMS-MS screening workflow. Data are first collected on an LC-IMS-MS platform. The LC dimension is then collapsed by summing the individual IMS-MS spectra across all retention times, resulting in a single IMS-MS spectrum for each sample. The summed spectra are imported into R where the m/z values and drift times are binned into consistent intervals. Data are subsequently split into training and testing sets at the sample level, and normalization and filtering are performed based on the distribution of the training data before being applied to the testing data. Following all data cleaning steps, relevant coordinates are selected from the training data using a bootstrapped Lasso, which are then used to create classification models that are evaluated on the testing data.