Table 3 Predictive performance of the BulkECexplorer

From: The BulkECexplorer compiles endothelial bulk transcriptomes to predict functional versus leaky transcription

  1. The table shows specificity and sensitivity metrics calculated for known EC-functional genes (actual positives) and known non-EC genes (actual negatives) with the GMM or zTPM classification models and thresholds of >0 TPM or >1 TPM. Each eligible dataset for each of the four classification approaches was treated as an individual test (n = 240 for transcript level >0 TPM or >1 TPM, n = 220 for zTPM, n = 198 for GMM). Original run: n = 37 actual positives and n = 109 actual negatives; refined run after subtracting six HUVEC proteome-expressed genes from the list of actual negatives: n = 37 actual positives and n = 103 actual negatives; amended run after subtracting markers of neural, glial, mural and myeloid cells from the list of actual negatives: n = 37 actual positives and n = 49 actual negatives.
Back to article page