The TME in patients receiving flotetuzumab immunotherapy. (A) Unsupervised hierarchical clustering (Euclidean distance, complete linkage) of immune gene expression (n = 770 genes in the NanoString’s PanCancer IO360 panel) in baseline BM samples from 38 patients with R/R AML treated with flotetuzumab immunotherapy (color-coded per the legend). Complete responses were defined as either CR, CRh, CRi, or MLFS at the end of cycle 1. PRs were defined as >50% decrease in BM blasts from baseline or decrease to 5% to 25% BM blasts at the end of cycle 1. ClustVis, an online tool for clustering of multivariate data, was used for data analysis and visualization.⁴⁰  (B) Expression of the top 10 genes associated with complete response to flotetuzumab (CR, CRh, CRi). A ranked gene list (χ² values) was generated using the Orange3 software package (version 3.25.0). Unsupervised hierarchical clustering (Euclidean distance, complete linkage). The immune cluster was defined as previously detailed.⁹  (C) Heatmap summarizing the correlation coefficients (color-coded per the legend) between our 10-gene signature score and immune cell type–specific and biological activity signature scores in baseline BM samples from patients with R/R AML. (D) Analysis of functional protein association networks using STRING (https://string-db.org/). Top 20 molecules interacting with the top 10 genes in our signature are shown together with their predicted mode of action (highest confidence interaction scores > 0.900). Network nodes (query proteins) represent proteins produced by a single protein-coding gene locus. White nodes represent second shells of interactors. Empty and filled nodes indicate proteins of unknown or partially known 3-dimensional structure, respectively. Edges represent protein-protein associations. Line shapes denote predicted modes of action. (E) AUROC curve measuring the predictive ability of the 10-gene signature score for antileukemic activity from flotetuzumab. Standard errors and confidence intervals are provided in supplemental Table 11. AUROC = 1.0, perfect prediction; AUROC = 0.5, no predictive ability.