2CALM workflow.

(a) represents the features of the 2CALM platform. Datasets are pre-processed using drift correction and filtering of outliers (optionally: ROI-extraction). For a full image comparison, dataset sizes are optionally equalized. The default bootstrapping step re-samples the datasets into sub-samples (used for spatial clustering). Hierarchical spatial clustering enables the extraction of the density-distribution and the curvature-distribution of clusters. For comparison, a two-level, multi-parameter analysis of different clustering dimensions is performed. Kolmogorov-Smirnov-, Wilcoxon-test and bootstrap confidence interval analysis is used for a comparison of protein distributions. (b) depicts the structure of the multilayer perceptron (MLP) neural network based machine learning module. The clustering from (a) provides parameters (feature vectors) which are used to train a MLP (including one hidden layer) network to assign the datasets to their classes (similarity classification).