Single-Cell Proteomic Classifier for Elderly AML Treatment Response

Featuring single-cell network profiling (SCNP) data from bone marrow and peripheral blood samples of 145 patients over 55 years old with acute myeloid leukemia (AML). It was used to train and validate a diagnostic classifier (DXSCNP) for predicting response to induction chemotherapy, achieving an AUROC of 0.76 in a verification set. The classifier's performance was compared against a clinical-only model (DXCLINICAL2).

Use Cases

• Train a classifier to predict CR/CRi vs. RD using SCNP-derived cell survival, differentiation, and apoptosis pathway signaling features.
• Validate the prognostic power of the DXSCNP classifier against the DXCLINICAL2 model using AUROC metrics from the BM Validation Set.
• Analyze single-cell proteomic signaling data from bone marrow and peripheral blood samples to identify biomarkers independent of traditional prognostic markers.

Strengths

• Data is derived from 145 SCNP-evaluable patients across four SWOG AML trials, providing a multi-trial cohort.
• Classifier validation includes an independent BM verification sample set of 24 patients from ECOG AML trials.
• The final DXSCNP classifier achieved statistically significant AUROC scores (0.76 and 0.72) in independent test sets.

Limitations

• The dataset is specific to elderly patients (>55 years old) with non-M3 AML, limiting generalizability to other age groups or AML subtypes.
• Sample sizes for the Training (N=74) and Validation (N=71) sets are moderate, which may constrain the complexity of models that can be robustly trained.
• The raw column structure and specific SCNP measurements are not detailed, requiring domain expertise for interpretation and reuse.

Provenance

Source

Data originates from four SWOG AML trials and an independent verification set from ECOG AML trials.

Collection Method

Single-cell network profiling (SCNP) data generated from multi-parametric flow cytometry analysis of patient samples.

Freshness

Last updated in June 2020.