Supplementary Figures 1-8, Supplementary Tables 1-6 from The Initial Detection and Partial Characterization of Circulating Tumor Cells in Neuroendocrine Prostate Cancer

Figure S1: Examples of a decision boundary that separate class A (red) from class B (blue) in 1 dimension (figure 1A) and 2 dimensions (figure 1B); Figure S2: Schematic of the supervised learning process (A) and leave one out cross validation (B); Figure S3: Single iteration of Leave-One-Out Cross-Validation performed at the blood sample level; Figure S4: An example of concordance of AURKA amplification in tumor and CTCs; Figure S5: (A) Kernel Density Estimate (KDE) curves of the classifier output are plotted for each patient sample colored by their diagnosis: NEPC (red) and CRPC (blue); Figure S6: Cell-level NEPC classifier (A) Receiver-Operating-Characteristic (ROC) curve generated on the classifier's single-cell output after LOOCV; Figure S7: To address whether the CTC classifier is simply stochastic, a reflection of an overall higher CTC count, linearity was assessed with a Pearson's coefficient showing a weak relationship between frequency of NEPC CTCs and total cell count; Table S1: A summary of cell-level features utilized to train Random Forest cell-level classifiers in both the LOOCV and for the classification of CTCs in the test cohort; Table S2: Patient characteristics (discovery cohort) including prior systemic therapies, serum markers including PSA (ng/ml), Chromogranin (ng/ml), NSE (ng/ml) and CTC counts; Table S3: Liver metastases in NEPC vs. CRPC; Table S4: The median concentration of CK-negative and AR-negative CTCs in CRPC, atypical CRPC, and NEPC patients; Table S5: Confusion matrix for the ability of CD56 staining to discriminate patients diagnosed with Small Cell Carcinoma NEPC vs CRPC; Table S6: Example of results from a single tube of blood.