Fig 1.
Single Cell Network Profiling (SCNP) Technology.
Bone marrow or blood cells (1) are modulated, fixed and permeabilized (2), then stained with an antibody cocktail containing antibodies directed against both cell surface markers as well as post-translational modifications of intra-cellular proteins(3). Cells are acquired using multiparametric flow cytometry (4) thus allowing quantification of intracellular pathway activity in cell subsets identified by gating on lineage surface markers (5). Various metrics to quantify basal and induced signaling and to assess association with biologic and clinical outcomes are applied.
Fig 2.
A flow diagram showing all patients enrolled onto the SWOG parent AML trials and rationale for exclusion of patients from the final analysis sets. Text boxes describe the characteristics of patients carried forward.
Fig 3.
A flow diagram showing all patients enrolled onto the parent ECOG AML trials and rationale for exclusion of patients from the final analysis sets. Text boxes describe the characteristics of patients carried forward.
Fig 4.
Flowchart of the study design with descriptive schematics of the patient sets in the Training, Verification and Validation analysis sets. SWOG samples were randomized into a Training and a Validation Analysis set and were sorted by tissue type (PB or BM). An initial subset of classifiers was trained separately in PB and BM samples in the Training Analysis sets and then PB classifiers were applied to BM and BM classifiers were applied to PB. From this training process 5 candidate classifiers were selected and applied to the ECOG Verification Analysis set. The final SCNP classifier was further refined and applied to 1) ECOG Verification Analysis set, 2) SWOG BM Validation Analysis set and 3) SWOG PB Validation Analysis set.
Fig 5.
Pathways investigated using SCNP in the Training study.
Schematic of the cell signaling pathways probed in the Training set. An SCNP node consists of the combination of a modulator and the corresponding intracellular readout. Modulators are shown outside the cell initiating signaling pathways that produce an intracellular proteomic response (readouts shown below the curve indicating cell membrane).
Fig 6.
Gating: Identification of Blast Cell Population.
Illustration of gating to identify blast cell population and cPARP negative blast cells. Intact cells were identified using scatter. Amine Aqua was then used to identify viable cells. CD45 was then used to identify Blast Cells. In wells where short term signaling was assayed, the blast cells were gated using cPARP expression to identify healthy leukemic cells.
Table 1.
Candidate Models.
Table 2.
Locked DXSCNP Classifier Inputs.
Table 3.
Patient/Disease Characteristics of the BM Training, Verification and Validation Analysis Sets.
Table 4.
Patient/Clinical Characteristics of the PB Training and Validation Analysis Sets.
Table 5.
Baseline Characteristics for SWOG Patients: SCNP-Evaluable vs. SCNP-Nonevaluable.
Table 6.
Baseline Characteristics for ECOG Patients: SCNP-Evaluable vs. SCNP-Nonevaluable.
Fig 7.
Performance of Classifier in Subgroups (BM).
Prediction accuracy of DXSCNP in various subgroups in the BM Validation Analysis Set. For age and WBC, the subgroups were defined by thresholding at the median value. For all samples cytogenetic risk was determined using NCCN 2013 guideline criteria. Similarly to what is done in clinical practice, patients with unknown cytogenetics were imputed as intermediate risk cytogenetics. The point estimate of accuracy measured by AUROC and confidence intervals (Delong method) are also shown.
Fig 8.
Comparison of predictions between paired PB and BM samples.
Predicted probability of CR for BM and PB samples from donors with SCNP data with paired samples in the validation set. Denovo vs secondary AML subtypes are noted in the inset. A majority of the predictions were concordant between the tissues types of de novo. Of note, two RD donors that were discordant are secondary AML.
Table 7.
Prediction accuracy of DXSCNP in BM and PB Validation Analysis subsets defined by AML Type (De Novo vs. Secondary) and availability of samples for both tissue types.