Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2022 Nov 11;23(22):13919.
doi: 10.3390/ijms232213919.

DRPreter: Interpretable Anticancer Drug Response Prediction Using Knowledge-Guided Graph Neural Networks and Transformer

Jihye Shin  1 , Yinhua Piao  2 , Dongmin Bang  1   3 , Sun Kim  1   2   4   5 , Kyuri Jo  6
Affiliations

DRPreter: Interpretable Anticancer Drug Response Prediction Using Knowledge-Guided Graph Neural Networks and Transformer

Jihye Shin et al. Int J Mol Sci. .

Abstract

Some of the recent studies on drug sensitivity prediction have applied graph neural networks to leverage prior knowledge on the drug structure or gene network, and other studies have focused on the interpretability of the model to delineate the mechanism governing the drug response. However, it is crucial to make a prediction model that is both knowledge-guided and interpretable, so that the prediction accuracy is improved and practical use of the model can be enhanced. We propose an interpretable model called DRPreter (drug response predictor and interpreter) that predicts the anticancer drug response. DRPreter learns cell line and drug information with graph neural networks; the cell-line graph is further divided into multiple subgraphs with domain knowledge on biological pathways. A type-aware transformer in DRPreter helps detect relationships between pathways and a drug, highlighting important pathways that are involved in the drug response. Extensive experiments on the GDSC (Genomics of Drug Sensitivity and Cancer) dataset demonstrate that the proposed method outperforms state-of-the-art graph-based models for drug response prediction. In addition, DRPreter detected putative key genes and pathways for specific drug-cell-line pairs with supporting evidence in the literature, implying that our model can help interpret the mechanism of action of the drug.

Keywords: Explainable AI; artificial intelligence; cancer; drug discovery; drug sensitivity; graph neural networks; human health; pharmacogenomics; precision medicine; transcriptomics.

PubMed Disclaimer

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Figure 1
Figure 1
An overview of DRPreter. In the graph representation sections, embeddings of pathway subgraphs and drug molecule were obtained using GNN. With the obtained pathway embeddings and drug embeddings as inputs to the transformer-based cell-line and drug fusion module, the embeddings were updated by reflecting inter-pathway relationships and pathway-drug relationships in the model learning process.
Figure 2
Figure 2
Box plot of drug-specific IC50 distributions of cell lines. The distribution of GDSC2 data (blue) compared with predicted missing IC50 values (orange). The 10 drugs with the highest median IC50 values and the 10 drugs with the lowest median were selected. Among the 20 drugs, IC50 value distributions of 18 drugs showed no significant differences through the Mann–Whitney Wilcoxon Test. ns: not significant, *: 0.01 < p-value < 0.05.
Figure 3
Figure 3
A detailed structure of type-aware transformer encoder reflecting interactions and relationships between pathways and a drug. We extracted drug-pathway interaction information from the modified encoder of the Transformer module and identified putative key pathways for the drug’s mechanism of action using a matrix of self-attention scores between pathways and the drug.
Figure 4
Figure 4
Visualization of all-pairwise self-attention scores from the transformer. (a) Dasatinib and leukemia cell line MEG-01 pair. (b) Dasatinib and breast cancer cell line BT-483. The figures show the y-axis as the query of the transformer and the x-axis as the key. On each axis, there is a drug and 34 pathways which start with “hsa”, indicating KEGG pathway identifiers.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Kellogg R.A., Dunn J., Snyder M.P. Personal omics for precision health. Circ. Res. 2018;122:1169–1171. doi: 10.1161/CIRCRESAHA.117.310909. - DOI - PubMed
    1. Ahmed Z. Practicing precision medicine with intelligently integrative clinical and multi-omics data analysis. Hum. Genom. 2020;14:1–5. doi: 10.1186/s40246-020-00287-z. - DOI - PMC - PubMed
    1. Kalamara A., Tobalina L., Saez-Rodriguez J. How to find the right drug for each patient? Advances and challenges in pharmacogenomics. Curr. Opin. Syst. Biol. 2018;10:53–62. doi: 10.1016/j.coisb.2018.07.001. - DOI - PMC - PubMed
    1. Singh D.B. The impact of pharmacogenomics in personalized medicine. Curr. Appl. Pharm. Biotechnol. 2019;171:369–394. - PubMed
    1. Cho S.Y. Patient-derived xenografts as compatible models for precision oncology. Lab. Anim. Res. 2020;36:1–11. doi: 10.1186/s42826-020-00045-1. - DOI - PMC - PubMed

MeSH terms

Substances

LinkOut - more resources