Isomarker: An Explainable Machine Learning Framework for Identification of Cell State Markers in Acute Myeloid Leukemia
Bibliographic record
Abstract
Acute myeloid leukemia (AML) is a hematologic malignancy classically associated with an expansion of poorly differentiated myeloid cells. Single-cell RNA-sequencing (scRNA-seq) studies have advanced our understanding of differentiation arrest across the leukemia cell hierarchy and their relation to prognosis, relapse, and drug response. Leukemic cellular hierarchy composition, such as Primitive versus GMP or Primitive versus Mature axes of cell state abundance, can define response to chemotherapy or targeted therapy sensitivity respectively (Zeng et al. Nature Med 2022). However, it remains unclear how these transcriptional AML cell states intersect with existing biomarkers used for clinical characterization of AML, such as cell surface markers used by clinical flow cytometry. This study aims to nominate existing and new surface markers associated with transcriptionally-defined AML cell states across leukemia cell hierarchies. We developed a novel consensus marker identification approach, IsoMarker, to identify a minimal set of enriched (differential expression), specific (auROC), and predictive (SHAP) cell surface markers for each AML cell state. To validate the utility of this approach, we generated an integrated scRNA-seq atlas of 1,079,555 single cells spanning 316 AML patient samples and 21 datasets, identifying 13 consensus leukemia cell states. Using IsoMarker, we identified 56 positive and 33 negative cell surface markers which effectively discern between the 13 leukemia cell states. We validated these markers by confirming that the expression levels of each positive marker were positively correlated with transcriptionally-inferred cell state abundance from 1037 total patients across the TCGA, BEAT-AML, and Leucegene cohorts. Composition of different leukemic cell states is known to govern functional leukemic stem cell activity, leading to heterogeneity in response to therapy and overall survival of AML patients. To extend the diagnostic utility of our nominated cell-state markers to include prediction of clinical outcomes, we conducted an integrative analysis of the positive AML cell state markers in relation to leukemic stem cell activity, ex vivo drug response, and meta-analysis of overall survival. Among the candidate positive cell state markers, 10 markers were associated with functional leukemia stem cell engraftment in mice, 5 markers were associated with the GMP cell state that defines response to chemotherapy, and 4 markers were associated with overall survival outcomes. We conducted a literature review of the 56 candidate positive markers in the context of AML to determine if IsoMarker can resolve both well-known and novel markers of AML cell states. Out of the 59 positive markers identified by IsoMarker, 21 have been previously reported as diagnostic markers of AML cell states. There are 35 positive markers that have not been characterized in the literature review and may serve as candidate, context-specific biomarkers of AML cell states. In summary, this study aims to translate insights from single-cell transcriptomics into hematopathology practice by advancing identification of markers that discern leukemic cell states and predict clinical outcomes. Applied to the task of identifying markers of AML cell states, IsoMarker is able to both resolve well-known markers and nominate novel markers useful for characterization of AML in relation to prognosis, relapse, and drug response. soMarker is a powerful framework for marker identification of cell states in single cell studies that can be generalized to other types of cancer.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".