ATLAS-AML: An Automated Bioinformatics Pipeline for Drug Target Characterization in Acute Myeloid Leukemia
Bibliographic record
Abstract
Acute myeloid leukemia (AML) is an aggressively heterogeneous disease with poor survival outcomes. An important checkpoint for AML drug development is ensuring target expression is enriched on leukemia cells compared to normal hematopoietic cells to avoid perturbing normal hematopoiesis. However, drugs satisfying this criterion still face additional challenges attributable to heterogeneity in AML, causing patients to develop resistance and relapse. In AML, chemoresistance and relapse are mediated by adverse genomic drivers and leukemia stem cell (LSC)-enriched cellular hierarchies (Zeng Nat Med 2022). This underscores the importance of identifying targets both enriched in leukemia and associated with sources of AML heterogeneity. Although many transcriptomic tools and pipelines have emerged for AML, none have linked gene expression to deep functional properties and non-genetic sources of intersample heterogeneity to enable informed predictions. To bridge this gap, we introduce ATLAS-AML, an automated bioinformatics pipeline for transcriptomic meta-analysis of genes and gene signatures in adult and pediatric AML. ATLAS-AML integrates 30 bulk (2172 donors) and single-cell (975,220 cells from 283 donors) RNA-sequencing datasets with preconfigured pipelines to streamline three key analyses for genes or gene signatures: (1) expression across normal and leukemic hematopoietic hierarchies, (2) enrichment in functionally-validated LSC+ fractions and (3) associations with relapse, genomics, clinical characteristics and patient survival. For established targets, ATLAS-AML can determine which patients, based on their mutational and cytogenetic profiles, are most likely to respond favorably to a potential treatment. ATLAS-AML is available as a containerized R package that experimental scientists can employ without bioinformatics expertise. To demonstrate how ATLAS-AML can guide target prioritization and characterization, we systematically analyzed published targets in ATLAS-AML. We identified 70 targets reported in the literature to be either enriched in leukemia, associated with LSCs, overexpressed at relapse, or predictive of patient survival. After benchmarking each gene against the aforementioned outcomes in ATLAS-AML, we observed that published targets were often optimized for particular outcomes, potentially overlooking other critical perspectives of AML biology. For example, several genes were enriched on leukemia cells, but not associated with genomics and stemness perspectives of AML heterogeneity nor clinical outcomes like relapse and patient survival. ATLAS-AML also constitutes a powerful framework for accelerating new target discovery. Reinterrogating our dataset with differential expression, we identified genes with enrichment in the same outcomes we benchmarked published targets against. Applying a meta p-value analysis, ATLAS-AML uncovered 13 targets that were overexpressed on leukemia cells compared to normal hematopoietic cells, enriched in functionally-validated LSCs, associated with disease relapse, and predictive of patient survival. For example, ATLAS-AML nominated CNST, a trans-Golgi network receptor for targeting connexins to the plasma membrane. CNST is differentially expressed on leukemia cells (p=0.000043), linked to functional LSC engraftment (p=0.000015), overexpressed at relapse (p=0.00036) and associated with poor prognosis in multivariable survival analysis in three independent cohorts (HR 1.23, p=0.04), lending to CNST's therapeutic viability in AML. Furthermore, ATLAS-AML determined that CNST expression was highest in RUNX1-mutated AMLs, identifying a patient group to prioritize for anti-CNST therapies. Altogether, ATLAS-AML enables scientists to leverage insights from single-cell and bulk transcriptomics to inform preclinical studies towards risk-tailored treatments in AML.
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 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 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".