ASGARD: A Single-cell Guided pipeline to Aid Repurposing of Drugs

Bing He, Lana X. Garmire, Liang, Haodong, Huang, Qianhui, Du, Yuheng, Li, Yijun, Garmire, David, Sun, Duxin, Garmire, Lana X.PubMed 2021 · DOI · arXiv

Abstract

Intercellular heterogeneity is a major obstacle to successful precision medicine. Single-cell RNA sequencing (scRNA-seq) 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 new drug recommendation system called: A Single-cell Guided Pipeline to Aid Repurposing of Drugs (ASGARD). ASGARD defines a novel drug score predicting drugs by considering all cell clusters to address the intercellular heterogeneity within each patient. We tested ASGARD on multiple diseases, including breast cancer, acute lymphoblastic leukemia, and coronavirus disease 2019 (COVID-19). On single-drug therapy, ASGARD shows significantly better average accuracy (AUC of 0.92) compared to two other bulk-cell-based drug repurposing methods (AUC of 0.80 and 0.76). It is also considerably better (AUC of 0.82) than other cell cluster level predicting methods (AUC of 0.67 and 0.55). In addition, ASGARD is also validated by the drug response prediction method TRANSACT with Triple-Negative-Breast-Cancer patient samples. Many top-ranked drugs are either approved by FDA or in clinical trials treating corresponding diseases. In silico cell-type specific drop-out experiments using triple-negative breast cancers show the importance of T cells in the tumor microenvironment in affecting drug predictions. In conclusion, ASGARD is a promising drug repurposing recommendation tool guided by single-cell RNA-seq for personalized medicine.

Introduction

The "one-size-fits-all" approach to medicine often fails because diseases like cancer or severe COVID-19 are incredibly heterogeneous. A single tumor is not a uniform mass of identical cells; rather, it is a complex ecosystem of malignant cells, immune cells, and structural cells interacting within a microenvironment.

Traditional drug repurposing methods typically rely on bulk RNA sequencing, which measures the average gene expression of a sample. This "smoothie" approach obscures critical information: a drug might effectively target the primary tumor cells while leaving a resistant sub-population untouched, or it might inadvertently suppress the immune cells necessary to combat the disease. To achieve true precision medicine, we require tools that can address the entire cellular landscape of an individual patient. To meet this need, we developed ASGARD (A Single-cell Guided Pipeline to Aid Repurposing of Drugs).

The ASGARD Framework

ASGARD leverages the high resolution of single-cell RNA sequencing (scRNA-seq) to identify drugs that can address the complex cellular composition of a patient. The workflow consists of four primary stages:

  1. Single-Cell Mapping: The pipeline uses scRNA-seq to identify and cluster every cell type within a patient sample.
  2. Differential Expression Pairing: ASGARD pairs diseased cell clusters with healthy "control" clusters using "anchor" genes. This allows the system to identify exactly how each specific cell type has deviated from its healthy state.
  3. Drug Matching: These disease-specific genetic signatures are compared against the LINCS L1000 database, which contains gene expression "fingerprints" for over 21,000 drugs and compounds.
  4. The Novel Drug Score: The core innovation of ASGARD is its comprehensive drug score. Instead of focusing on a single cell type, ASGARD calculates a score that weights a drug’s potential effectiveness based on the proportion of each cell type in the patient, the statistical significance of the drug's ability to reverse the disease signature, and the ratio of specific deregulated genes that the drug successfully "fixes."

Experimental Validation

We validated ASGARD across three distinct disease models to test its versatility and accuracy:

  • Breast Cancer: Utilizing data from Triple-Negative Breast Cancer (TNBC) patients and Patient-Derived Xenografts (PDX).
  • Leukemia: Analyzing pediatric bone marrow samples from patients with Pre-T acute lymphoblastic leukemia.
  • COVID-19: Examining single-cell data from the bronchoalveolar lavage fluid (BALF) of 15 severe patients, comparing those who recovered against those who did not.

In single-drug therapy simulations, ASGARD achieved an average AUC of 0.92. This significantly outperforms traditional bulk-cell methods (which reached AUCs of 0.76–0.80) and previous single-cell cluster-level methods (which ranged between 0.55 and 0.67). Furthermore, the pipeline demonstrated high robustness, maintaining accuracy even when the number of cells was artificially reduced or the proportions of diseased cells were altered in simulations.

Key Insights

The Importance of the Microenvironment

ASGARD’s multi-cluster drug score proved that targeting a single cell cluster is insufficient. A holistic view of the tissue is required for effective repurposing. Notably, in "drop-one-out" experiments where specific cell types were removed from the analysis, the removal of T cells in breast cancer samples caused the most significant shifts in drug recommendations. This highlights the critical role the immune microenvironment plays in determining therapeutic success.

Clinical Alignment and New Candidates

Many of ASGARD’s top-ranked hits are already FDA-approved for the diseases studied, such as Neratinib for breast cancer and Tretinoin for leukemia, providing a strong validation of the model's predictive power. Additionally, the pipeline identified promising repurposing candidates:

  • Mebendazole: A common anti-parasitic medication was flagged as a top candidate for TNBC, aligning with recent clinical interest in its anti-cancer properties.
  • ACE Inhibitors: For severe COVID-19, ASGARD identified Rescinnamine and Enalapril. Given that SARS-CoV-2 enters cells via the ACE2 receptor, finding these drugs through a purely genetic-signature-based approach provides strong computational support for their potential utility.

Limitations and Future Directions

While ASGARD represents a significant advancement, some limitations remain. The drug reference data (LINCS L1000) is derived from lab-grown cell lines, which may not perfectly replicate drug behavior in primary human tissue. Additionally, identifying perfectly matched healthy control tissue for every patient remains a logistical challenge in clinical practice. Future iterations of the pipeline may benefit from tissue-specific drug references to further refine recommendation accuracy.

Conclusion

ASGARD is a promising drug repurposing tool that accounts for intercellular heterogeneity to provide personalized therapeutic recommendations. By moving beyond the average view of bulk sequencing, ASGARD offers a more nuanced and accurate approach to precision medicine. The tool is available for educational use at https://github.com/lanagarmire/ASGARD.

Summary updated

Abstract

Intercellular heterogeneity is a major obstacle to successful precision medicine. Single-cell RNA sequencing (scRNA-seq) 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 new drug recommendation system called: A Single-cell Guided Pipeline to Aid Repurposing of Drugs (ASGARD). ASGARD defines a novel drug score predicting drugs by considering all cell clusters to address the intercellular heterogeneity within each patient. We tested ASGARD on multiple diseases, including breast cancer, acute lymphoblastic leukemia, and coronavirus disease 2019 (COVID-19). On single-drug therapy, ASGARD shows significantly better average accuracy (AUC of 0.92) compared to two other bulk-cell-based drug repurposing methods (AUC of 0.80 and 0.76). It is also considerably better (AUC of 0.82) than other cell cluster level predicting methods (AUC of 0.67 and 0.55). In addition, ASGARD is also validated by the drug response prediction method TRANSACT with Triple-Negative-Breast-Cancer patient samples. Many top-ranked drugs are either approved by FDA or in clinical trials treating corresponding diseases. In silico cell-type specific drop-out experiments using triple-negative breast cancers show the importance of T cells in the tumor microenvironment in affecting drug predictions. 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.
BibTeX
@inproceedings{he2021asgard,
  author = {Bing He and Lana X. Garmire and Liang, Haodong and Huang, Qianhui and Du, Yuheng and Li, Yijun and Garmire, David and Sun, Duxin and Garmire, Lana X.},
  title = {{ASGARD: A Single-cell Guided pipeline to Aid Repurposing of Drugs}},
  booktitle = {PubMed},
  year = {2021},
  doi = {10.48550/arxiv.2109.06377},
}
Figure