Bioinformatic screening to find AML-specific pathways. (A) Overall strategy for database screening. (B) Graphical illustration of 4 pathway clusters upregulated in AML BM samples from panel A, using GSCluster²⁶ R package. The number of connected gene sets in each cluster is indicated. (C) Comparison of UBA2 and SUMO1 gene expression between healthy and AML BM samples in OHSU and MILE databases. (D) Comparison of UBA2 and SUMO1 gene expression between healthy and AML BM/peripheral blood samples in Gene Expression Omnibus (GEO) data sets by Roushangar and Mias.²⁵ (E) Representative western blot for SAE2, SAE1, UBC9, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in peripheral blood from healthy controls and patients with AML at diagnosis or remission state after treatment (left). The intensities of the bands from all the samples were quantified by densitometry and displayed as the ratio of each protein to GAPDH (loading control) (right). Newly diagnosed patients with AML (n = 7), those at remission state (n = 5), and healthy controls (n = 5). Results are expressed as the mean ± standard error of the mean. For panels C-E, P values are from Wilcoxon rank-sum test; ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. BM, bone marrow; GEO, Gene Expression Omnibus; FDR, false discovery rate; GSEA, gene set enrichment analysis; rRNA, ribosomal RNA.