Student or educator? Sign up with your school email and get 6 months of Pro free.

The DOI matches its Crossref record; 81 of the 81 references it cites were found in a registry or an index; 31 of the 72 sources its claims cite were checked for support.

ASGARD is A Single-cell Guided Pipeline to Aid Repurposing of Drugs

Open access

Abstract

Single-cell RNA sequencing technology has enabled in-depth analysis of intercellular heterogeneity in various diseases. However, its full potential for precision medicine has yet to be reached. Towards this, we propose A Single-cell Guided Pipeline to Aid Repurposing of Drugs (ASGARD) that defines a drug score to recommend drugs by considering all cell clusters to address the intercellular heterogeneity within each patient. ASGARD shows significantly better average accuracy on single-drug therapy compared to two bulk-cell-based drug repurposing methods. We also demonstrated that it performs considerably better than other cell cluster-level predicting methods. In addition, we validate ASGARD using the drug response prediction method TRANSACT with Triple-Negative-Breast-Cancer patient samples. We find that many top-ranked drugs are either approved by the Food and Drug Administration or in clinical trials treating corresponding diseases. In conclusion, ASGARD is a promising drug repurposing recommendation tool guided by single-cell RNA-seq for personalized medicine. ASGARD is free for educational use at https://github.com/lanagarmire/ASGARD.

Overview

CoChat Summary

TL;DR

ASGARD is a single-cell RNA-seq-guided drug repurposing pipeline that defines a multi-cluster drug score to account for intercellular heterogeneity within patients. It reports higher drug/compound prediction AUCs than bulk-based and other single-cell-based repurposing methods across breast cancer, leukemia, and COVID-19 datasets.

Key findings

  • ASGARD’s multi-cluster drug/compound AUCs were 0.92 for breast cancer, 0.95 for leukemia, and 0.88 for COVID-19, compared with CLUE (0.74, 0.82, 0.85) and DrInsight (0.81, 0.73, 0.73).
  • At the single-cluster level, ASGARD had higher average AUCs than Alakwaa’s and Guo’s pipelines: breast cancer 0.83 vs 0.76 and 0.54; leukemia 0.81 vs 0.51 and 0.52; COVID-19 0.82 vs 0.72 and 0.58.
  • In applications, ASGARD predicted 11 drugs for advanced metastatic breast cancer PDX models, with fostamatinib top-ranked and colchicine second; in TNBC patient samples it predicted 13 drugs, including mebendazole, crizotinib, neratinib, and vinblastine overlapping with PDX predictions.
  • For Pre-T ALL, the top candidate was tretinoin, an FDA-approved leukemia treatment; for severe COVID-19, rescinnamine (2nd) and enalapril (4th) were predicted, both ACE inhibitors.

Methods

  • ASGARD pairs diseased and normal cell clusters using “anchor” genes, imports differentially expressed genes (adjusted P-value < 0.05) from user-selected methods such as Limma, Seurat, DESeq2, or edgeR, and identifies drugs reversing expression patterns using LINCS L1000 drug response data, DrInsight, outlier-sum statistics, and Kolmogorov–Smirnov tests with Benjamini–Hochberg FDR adjustment.
  • The drug score combines cell type proportion, single-cluster FDR significance, and the ratio of reversed significantly deregulated genes across selected clusters; ASGARD also provides a standardized drug score and Fisher’s combined P-value.
  • Data included GEO scRNA-seq datasets: GSE161529 (4 TNBC patients, 4 controls), GSE123926 and GSE113197 (advanced metastatic TNBC PDX and normal breast epithelial cells), GSE132509 (2 Pre-T ALL patients, 3 controls), and GSE145926/GSE158055 (15 severe COVID-19 patients: 4 deceased, 11 cured). Evaluation used ROC/AUC with FDA-approved drugs and advanced clinical/animal-validated compounds as positives, plus simulation data based on GSE123926 and GSE113197.

Limitations

  • The paper states that the best controls are normal tissues from the same patients, followed by matched patients without the disease, but such normal samples may be difficult to obtain; early-stage disease samples or tissues of relevant origin may be substitutes.
  • The drug reference uses LINCS L1000 responses collected from 98 cell lines; the authors recommend tissue-specific drug references and, if unavailable, identifying a cell line whose baseline expression is most similar to control samples after adjusting for cell line vs. primary tissue differences.
  • The paper notes sparsity and heterogeneity in single-cell data can cause false discoveries of differentially expressed genes, and DE method choice affected downstream AUC performance. Evident from the text: validation is largely computational/in silico and retrospective, with TRANSACT comparison limited to drugs in the GDSC dataset.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41467-023-36637-3

Cites

81 of 81 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 The workflow of the ASGARD drug repurposing pipeline. The workflow of the ASGARD pipeline. Diseased and normal cells are paired according to “anchor” genes that are expressed consistently between the two types of cells. The differentially expressed (DE) genes are identified between diseased and normal cells, either within a cluster or within a cell type. Using the consistent DE genes as the input, potential drugs that significantly reverse the pattern of DE genes are identified, using the Kolmogorov–Smirnov (K-S) test with Benjamini–Hochberg (BH) false discovery rate (FDR) adjustment. Next ASGARD estimates and ranks the drug scores for single drugs, by targeting specific cell cluster(s) or all cell clusters.
Fig. 1 The workflow of the ASGARD drug repurposing pipeline. The workflow of the ASGARD pipeline. Diseased and normal cells are paired according to “anchor” genes that are expressed consistently between the two types of cells. The differentially expressed (DE) genes are identified between diseased and normal cells, either within a cluster or within a cell type. Using the consistent DE genes as the input, potential drugs that significantly reverse the pattern of DE genes are identified, using the Kolmogorov–Smirnov (K-S) test with Benjamini–Hochberg (BH) false discovery rate (FDR) adjustment. Next ASGARD estimates and ranks the drug scores for single drugs, by targeting specific cell cluster(s) or all cell clusters.CC BY · PubMed Central
Fig. 2 Comparing ASGARD to bulk-cell-based repurposing methods. The receiver operating characteristic (ROC) curves and area under curve (AUC) scores of the ASGARD, on advanced metastatic breast cancer, acute lymphoblastic leukemia, and coronavirus disease 2019 (COVID-19), respectively. a Comparison among different DE analysis methods Limma (red), DESeq2 (light blue), Seurat (green), and edgeR (blue) using the best-performing mode in each. b Comparison of ASGARD (red) and bulk-sample based drug repurposing methods: CLUE (blue), and DrInsight (green), using the same three diseases as in a. The single-cell RNA-Seq data were aggregated to pseudo-bulk RNA-Seq data as the input of the bulk-sample based methods. Source data are provided as a Source Data file.
Fig. 2 Comparing ASGARD to bulk-cell-based repurposing methods. The receiver operating characteristic (ROC) curves and area under curve (AUC) scores of the ASGARD, on advanced metastatic breast cancer, acute lymphoblastic leukemia, and coronavirus disease 2019 (COVID-19), respectively. a Comparison among different DE analysis methods Limma (red), DESeq2 (light blue), Seurat (green), and edgeR (blue) using the best-performing mode in each. b Comparison of ASGARD (red) and bulk-sample based drug repurposing methods: CLUE (blue), and DrInsight (green), using the same three diseases as in a. The single-cell RNA-Seq data were aggregated to pseudo-bulk RNA-Seq data as the input of the bulk-sample based methods. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 3 Comparing ASGARD to other single-cell-based repurposing methods. The receiver operating characteristic (ROC) curves and area under curve (AUC) scores of the ASGARD (red) and other published pipelines: Alakwaa’s pipeline (green) and Guo’s pipeline (blue). The results of drug/compound repurposing are shown on every cell cluster of the metastatic breast cancer dataset (a), every cell cluster of the acute lymphoblastic leukemia dataset (b) and 4 clusters with increased cell proportions in the decreased severe vs cured severe COVID-19 patients (c). The proportion of each single-cell cluster is shown in the brackets above each plot. Source data are provided as a Source Data file.
Fig. 3 Comparing ASGARD to other single-cell-based repurposing methods. The receiver operating characteristic (ROC) curves and area under curve (AUC) scores of the ASGARD (red) and other published pipelines: Alakwaa’s pipeline (green) and Guo’s pipeline (blue). The results of drug/compound repurposing are shown on every cell cluster of the metastatic breast cancer dataset (a), every cell cluster of the acute lymphoblastic leukemia dataset (b) and 4 clusters with increased cell proportions in the decreased severe vs cured severe COVID-19 patients (c). The proportion of each single-cell cluster is shown in the brackets above each plot. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 4 Drug repurposing in triple-negative breast cancer (TNBC) patient samples. a UMAP plots of single-cell data from four TNBC patient samples and four control samples from the same study. b The overall drug scores combining all TNBC samples as well as drug scores in each TNBC sample, among the top-ranked significant single drugs (FDR < 0.05). Gray color indicates a lack of significance for a particular drug in an individual. The four drugs in the red boxes overlapped with the top drugs predicted by ASGARD using PDX models of TNBC patients. c Comparison between TRANSACT drug sensitivity and ASGARD drug score. Data are presented as mean values ± /SD. n = 4 biologically independent samples. d Heatmap of the drug score changes in the cell-type-specific drop-one-out experiment, where each cell type in the tumor microenvironment is removed at a time. e Pathway enrichment analysis (TNBC vs. normal) for each cell cluster. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. f The top drug candidate mebendazole, its target genes, pathways, and single-cell clusters. Orange node: up-regulated gene (logFC>1 and adjusted P-value < 0.05). Blue node: down-regulated gene (logFC < −1 and adjusted P-value < 0.05). Orange solid edge: drug stimulates gene expression. Blue solid edge: drug inhibits gene expression. The width of the edge is proportional to the strength of the drug effect. Gray dotted edge: gene belonging to a pathway. Gray backward slash: pathway significant in a cell cluster. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. Source data are provided as a Source Data file.
Fig. 4 Drug repurposing in triple-negative breast cancer (TNBC) patient samples. a UMAP plots of single-cell data from four TNBC patient samples and four control samples from the same study. b The overall drug scores combining all TNBC samples as well as drug scores in each TNBC sample, among the top-ranked significant single drugs (FDR < 0.05). Gray color indicates a lack of significance for a particular drug in an individual. The four drugs in the red boxes overlapped with the top drugs predicted by ASGARD using PDX models of TNBC patients. c Comparison between TRANSACT drug sensitivity and ASGARD drug score. Data are presented as mean values ± /SD. n = 4 biologically independent samples. d Heatmap of the drug score changes in the cell-type-specific drop-one-out experiment, where each cell type in the tumor microenvironment is removed at a time. e Pathway enrichment analysis (TNBC vs. normal) for each cell cluster. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. f The top drug candidate mebendazole, its target genes, pathways, and single-cell clusters. Orange node: up-regulated gene (logFC>1 and adjusted P-value < 0.05). Blue node: down-regulated gene (logFC < −1 and adjusted P-value < 0.05). Orange solid edge: drug stimulates gene expression. Blue solid edge: drug inhibits gene expression. The width of the edge is proportional to the strength of the drug effect. Gray dotted edge: gene belonging to a pathway. Gray backward slash: pathway significant in a cell cluster. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 5 Drug repurposing for precursor T cell acute lymphoblastic leukemia (Pre-T ALL). a UMAP plots of all cells from 3 normal controls and 2 Pre-T ALL samples. b UMAP plots of T cell clusters from normal controls and Pre-T ALL samples. c Pathway enrichment analysis (leukemia vs normal) for each T cell cluster. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. d The overall drug score and drug score in each Pre-T ALL patient, among top-ranked significant drugs (FDR < 0.05). Drug approved for leukemia treatment by the FDA is tretinoin. e The drug candidate tretinoin, its target genes, pathways, and single-cell clusters. All labels and their annotations are the same as Fig. 4f. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. Source data are provided as a Source Data file.
Fig. 5 Drug repurposing for precursor T cell acute lymphoblastic leukemia (Pre-T ALL). a UMAP plots of all cells from 3 normal controls and 2 Pre-T ALL samples. b UMAP plots of T cell clusters from normal controls and Pre-T ALL samples. c Pathway enrichment analysis (leukemia vs normal) for each T cell cluster. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. d The overall drug score and drug score in each Pre-T ALL patient, among top-ranked significant drugs (FDR < 0.05). Drug approved for leukemia treatment by the FDA is tretinoin. e The drug candidate tretinoin, its target genes, pathways, and single-cell clusters. All labels and their annotations are the same as Fig. 4f. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 6 Drug repurposing for reducing mortality of severe COVID-19 patients. a Single-cell populations of bronchoalveolar immune cells in 11 cured and four deceased severe COVID-19 patients, respectively. b The proportions of cell type in (left) and log10 transformed fold changes in deceased over the cured state (right) of the single-cell populations in a. c Pathway enrichment analysis (deceased severe vs. cured severe) for neutrophil, NK cell, T cell, and monocyte, respectively. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. d The drug scores in the four deceased severe COVID-19 patients and all four patients among top-ranked significant drugs (FDR < 0.05). e The drug candidate rescinnamine and enalapril, their target genes and pathways in the monocyte, NK cell, T cell, and neutrophil, respectively, from severe COVID-19 patients. All labels and their annotations are the same as in Fig. 4f. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. Source data are provided as a Source Data file.
Fig. 6 Drug repurposing for reducing mortality of severe COVID-19 patients. a Single-cell populations of bronchoalveolar immune cells in 11 cured and four deceased severe COVID-19 patients, respectively. b The proportions of cell type in (left) and log10 transformed fold changes in deceased over the cured state (right) of the single-cell populations in a. c Pathway enrichment analysis (deceased severe vs. cured severe) for neutrophil, NK cell, T cell, and monocyte, respectively. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. d The drug scores in the four deceased severe COVID-19 patients and all four patients among top-ranked significant drugs (FDR < 0.05). e The drug candidate rescinnamine and enalapril, their target genes and pathways in the monocyte, NK cell, T cell, and neutrophil, respectively, from severe COVID-19 patients. All labels and their annotations are the same as in Fig. 4f. P-values were determined by two-sided Fisher’s exact test and were adjusted by BH FDR. Source data are provided as a Source Data file.CC BY · PubMed Central

Verification

Claims the paper makes

14 Valid · 2 Not valid · 26 Not checked · 81 Own results and interpretations

69 more claims · 2 with their cited sources checked

Create a free account to see all 123 claims and to verify the 41 references still open.

See all claims — free account
  • It may state two separate claims as one. 36
  • Probably not a substantial claim. 22
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

4 of 4

Citations

Cites

Show all 80 references

Research works printed but not linked: 1

  1. Mourragui, S. et al. Predicting patient response with models trained on cell lines and patient-derived xenografts by nonlinear transfer learning. Proc. Natl. Acad. Sci. 118, e2106682118 (2021) · DOI

Cited by

See all 62

Metadata

Authors
He, B., Xiao, Y., Liang, H., Huang, Q., Du, Y., Li, Y., Garmire, D., Sun, D., & Garmire, L. X.
Affiliations
University of Michigan (US)
Corresponding author
Garmire, Lana X.
Venue
Nature Communications
Published
2023-02-22
Volume
14
Issue
1
Pages
993-993
ISSN
2041-1723 (ISSN-L)
OpenAlex
W4321479839
Cited by
57
References
80
Citations by year
2026: 12 · 2025: 20 · 2024: 14 · 2023: 11
FWCI
8.16
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Institutes of Health
  • National Cancer Institute (T32CA140044)
  • National Institute of Environmental Health Sciences (LM012373, K01ES025434)
  • U.S. National Library of Medicine (R01LM012907, HD084633, LM012373, R01LM012373)
  • Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01HD084633)
Topics
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Gene expression and cancer classification
Keywords
drug repurposing, single-cell RNA sequencing, intercellular heterogeneity, cell clusters, drug recommendation, precision medicine, triple-negative breast cancer, drug response prediction
MeSH terms
Pharmaceutical Preparations, Humans, Precision Medicine, Drug Repositioning

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: ASGARD is A Single-cell Guided Pipeline to Aid Repurposing of Drugs. https://app.cochat.ai/papers/10.1038/s41467-023-36637-3
The paper
He, B., Xiao, Y., Liang, H., Huang, Q., Du, Y., Li, Y., Garmire, D., Sun, D., & Garmire, L. X. (2023). ASGARD is A Single-cell Guided Pipeline to Aid Repurposing of Drugs. Nature Communications. https://doi.org/10.1038/s41467-023-36637-3

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.