Heterogeneous graph transformer with multi-source feature extraction for circRNA-drug association prediction

Abstract

Circular RNAs (circRNAs) play key regulatory roles in disease onset and progression. Predicting circRNA-drug associations can reveal underlying mechanisms and guide targeted drug discovery, yet current methods struggle to integrate multi-source biological data and capture high-order semantic relationships, limiting accuracy. We propose MTHCD, a multi-source feature fusion and heterogeneous graph neural network framework for circRNA-drug association prediction. MTHCD uses a bidirectional gated recurrent unit (BiGRU) to extract sequence features from k-mer-encoded circRNA data and a graph convolutional network (GCN) to obtain molecular structure features from drug SMILES graphs. Similarity priors and multi-scale random walk position encodings are incorporated to build a heterogeneous graph enriched with node and edge attributes, which is processed by a heterogeneous graph transformer to model multi-type nodes and relations, capturing cross-modal high-order associations. In extensive five-fold and ten-fold cross-validation, MTHCD consistently outperforms state-of-the-art methods across multiple metrics, achieving superior accuracy and robustness. This efficient, scalable framework can accelerate therapeutic target identification and drug repurposing for circRNA-related research, supporting precision medicine and novel drug development.

Graphical abstract

Keywords

circRNA / BiGRU / multi-scale random walk / heterogeneous graph transformer / k-mer encoding.

References

[1] Sanger H L, Klotz G, Riesner D, Gross H J, Kleinschmidt A K . Viroids are single-stranded covalently closed circular RNA molecules existing as highly base-paired rod-like structures. Proceedings of the National Academy of Sciences of the United States of America, 1976, 73( 11): 3852–3856

[2] Hsu M T, Coca-Prados M . Electron microscopic evidence for the circular form of RNA in the cytoplasm of eukaryotic cells. Nature, 1979, 280( 5720): 339–340

[3] Chen L L . The expanding regulatory mechanisms and cellular functions of circular RNAs. Nature Reviews Molecular Cell Biology, 2020, 21( 8): 475–490

[4] Kristensen L S, Andersen M S, Stagsted L V W, Ebbesen K K, Hansen T B, Kjems J . The biogenesis, biology and characterization of circular RNAs. Nature Reviews Genetics, 2019, 20( 11): 675–691

[5] Ngo L H, Bert A G, Dredge B K, Williams T, Murphy V, Li W, Hamilton W B, Carey K T, Toubia J, Pillman K A, Liu D, Desogus J, Chao J A, Deans A J, Goodall G J, Wickramasinghe V O . Nuclear export of circular RNA. Nature, 2024, 627( 8002): 212–220

[6] Enuka Y, Lauriola M, Feldman M E, Sas-Chen A, Ulitsky I, Yarden Y . Circular RNAs are long-lived and display only minimal early alterations in response to a growth factor. Nucleic Acids Research, 2016, 44( 3): 1370–1383

[7] Yang Y, Fan X J, Mao M W, Song X W, Wu P, Zhang Y, Jin Y F, Yang Y, Chen L L, Wang Y, Wong C C, Xiao X S, Wang Z F . Extensive translation of circular RNAs driven by N⁶-methyladenosine. Cell Research, 2017, 27( 5): 626–641

[8] Wang Y, Wang Z . Efficient backsplicing produces translatable circular mRNAs. RNA, 2015, 21( 2): 172–179

[9] Fan X, Yang Y, Chen C, Wang Z . Pervasive translation of circular RNAs driven by short IRES-like elements. Nature Communications, 2022, 13( 1): 3751

[10] Jackson L A, Anderson E J, Rouphael N G, Roberts P C, Makhene M, . et al. An mRNA vaccine against SARS-CoV-2-preliminary report. The New England Journal of Medicine, 2020, 383( 20): 1920–1931

[11] Mulligan M J, Lyke K E, Kitchin N, Absalon J, Gurtman A, Lockhart S, Neuzil K, Raabe V, Bailey R, Swanson K A, Li P, Koury K, Kalina W, Cooper D, Fontes-Garfias C, Shi P Y, Türeci Ö, Tompkins K R, Walsh E E, Frenck R, Falsey A R, Dormitzer P R, Gruber W C, Şahin U, Jansen K U . Publisher correction: phase I/II study of COVID-19 RNA vaccine BNT162b1 in adults. Nature, 2021, 590( 7844): E26

[12] Cai Z, Lu C, He J, Liu L, Zou Y, Zhang Z, Zhu Z, Ge X, Wu A, Jiang T, Zheng H, Peng Y . Identification and characterization of circRNAs encoded by MERS-CoV, SARS-CoV-1 and SARS-CoV-2. Briefings in Bioinformatics, 2021, 22( 2): 1297–1308

[13] Cui X, Qu X, Li D, Yang Y, Li Y, Zhang X . MKGCN: multi-modal knowledge graph convolutional network for music recommender systems. Electronics, 2023, 12( 12): 2688

[14] Ojo D, Kapoor A, Hill B, Lin X, Wong N, Pinthus J, Tang D . RKIP is commonly downregulated in clear cell renal cell carcinoma (ccRCC). Cancer Cell & Microenvironment, 2015, 5: e445

[15] Ji T T, Chen Q N, Tao S D, Shi Y Y, Chen Y, Shen L, Wang C L, Yu L . The research progress of circular RNAs in hematological malignancies. Hematology, 2019, 24( 1): 727–731

[16] Zhao X, Cai Y, Xu J . Circular RNAs: biogenesis, mechanism, and function in human cancers. International Journal of Molecular Sciences, 2019, 20( 16): 3926

[17] Guo Y, Lei X, Pan Y . An encoding-decoding framework based on CNN for circRNA-RBP binding sites prediction. Chinese Journal of Electronics, 2024, 33( 1): 256–263

[18] Guo Y, Lei X, Liu L, Pan Y . circ2CBA: prediction of circRNA-RBP binding sites combining deep learning and attention mechanism. Frontiers of Computer Science, 2023, 17( 5): 175904

[19] Ma M, Lei X . A dual graph neural network for drug-drug interactions prediction based on molecular structure and interactions. PLoS Computational Biology, 2023, 19( 1): e1010812

[20] Yang J, Lei X, Pan Y . Predicting circRNA-disease associations by using multi-biomolecular networks based on variational graph auto-encoder with attention mechanism. Chinese Journal of Electronics, 2024, 33( 6): 1526–1537

[21] Lei X, Chen Y, Pan Y . Multi-source data with Laplacian eigenmaps and denoising autoencoder for predicting microbe-disease associations via convolutional neural network. Journal of Computer Science and Technology, 2025, 40( 2): 588–604

[22] Deng L, Liu Z, Qian Y, Zhang J . Predicting circRNA-drug sensitivity associations via graph attention auto-encoder. BMC Bioinformatics, 2022, 23( 1): 160

[23] Yang B, Chen H . Predicting circRNA-drug sensitivity associations by learning multimodal networks using graph auto-encoders and attention mechanism. Briefings in Bioinformatics, 2023, 24( 1): bbac596

[24] Li G, Li Y, Liang C, Luo J . DeepWalk-aware graph attention networks with CNN for circRNA-drug sensitivity association identification. Briefings in Functional Genomics, 2024, 23( 4): 418–428

[25] Niu M, Wang C, Chen Y, Zou Q, Luo X . Interpretable multi-instance heterogeneous graph network learning modelling CircRNA-drug sensitivity association prediction. BMC Biology, 2025, 23( 1): 131

[26] Luo Y, Deng L . DPMGCDA: deciphering circRNA-drug sensitivity associations with dual perspective learning and path-masked graph autoencoder. Journal of Chemical Information and Modeling, 2024, 64( 10): 4359–4372

[27] Huang Z, Chen K, Xiao X, Fan Z, Zhang Y, Deng L . DeepHeteroCDA: circRNA-drug sensitivity associations prediction via multi-scale heterogeneous network and graph attention mechanism. Briefings in Bioinformatics, 2025, 26( 2): bbaf159

[28] Ruan H, Xiang Y, Ko J, Li S, Jing Y, Zhu X, Ye Y, Zhang Z, Mills T, Feng J, Liu C J, Jing J, Cao J, Zhou B, Wang L, Zhou Y, Lin C, Guo A Y, Chen X, Diao L, Li W, Chen Z, He X, Mills G B, Blackburn M R, Han L . Comprehensive characterization of circular RNAs in ~ 1000 human cancer cell lines. Genome Medicine, 2019, 11( 1): 55

[29] Yang W, Soares J, Greninger P, Edelman E J, Lightfoot H, Forbes S, Bindal N, Beare D, Smith J A, Thompson I R, Ramaswamy S, Futreal P A, Haber D A, Stratton M R, Benes C, McDermott U, Garnett M J . Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells. Nucleic Acids Research, 2013, 41( D1): D955–D961

[30] Weininger D . SMILES, a chemical language and information system. 1. Introduction to methodology and encoding rules. Journal of Chemical Information and Computer, 1988, 28( 1): 31–36

[31] Knox C, Wilson M, Klinger C M, Franklin M, Oler E, . et al. DrugBank 6. 0: the DrugBank knowledgebase for 2024. Nucleic Acids Research, 2024, 52( D1): D1265–D1275

[32] Rangwala S H, Kuznetsov A, Ananiev V, Asztalos A, Borodin E, Evgeniev V, Joukov V, Lotov V, Pannu R, Rudnev D, Shkeda A, Weitz E M, Schneider V A . Accessing NCBI data using the NCBI Sequence Viewer and Genome Data Viewer (GDV). Genome Research, 2021, 31( 1): 159–169

[33] Wang Y, Bryant S H, Cheng T, Wang J, Gindulyte A, Shoemaker B A, Thiessen P A, He S, Zhang J . PubChem BioAssay: 2017 update. Nucleic Acids Research, 2017, 45( D1): D955–D963

[34] Landrum G. RDKit: open-source cheminformatics from machine learning to chemical registration. Abstracts of Papers of the American Chemical Society, 2019, 258

[35] Liu Y, Ren X, Li J, Chen X, Zhu X . Prediction of circRNA-drug sensitivity using random auto-encoders and multi-layer heterogeneous graph transformers. Applied Intelligence, 2025, 55( 4): 238

[36] Rees M G, Seashore-Ludlow B, Cheah J H, Adams D J, Price E V, Gill S, Javaid S, Coletti M E, Jones V L, Bodycombe N E, Soule C K, Alexander B, Li A, Montgomery P, Kotz J D, Hon C S Y, Munoz B, Liefeld T, Dančík V, Haber D A, Clish C B, Bittker J A, Palmer M, Wagner B K, Clemons P A, Shamji A F, Schreiber S L . Correlating chemical sensitivity and basal gene expression reveals mechanism of action. Nature Chemical Biology, 2016, 12( 2): 109–116

[37] More L A, Lane S, Asnani A. 5-FU cardiotoxicity: vasospasm, myocarditis, and sudden death. Current Cardiology Reports, 2021, 23(3): 17

[38] Jin Z, Peng F, Zhang C, Tao S, Xu D, Zhu Z . Expression, regulating mechanism and therapeutic target of KIF20A in multiple cancer. Heliyon, 2023, 9( 2): e13195

[39] Zhao Y, He J, Li Y, Lv S, Cui H. NUSAP1 potentiates chemoresistance in glioblastoma through its SAP domain to stabilize ATR. Signal Transduction and Targeted Therapy, 2020, 5(1): 44

[40] Grant S, Easley C, Kirkpatrick P. Vorinostat. Nature Reviews Drug Discovery, 2007, 6(1): 21−22

[41] Banerjee N S, Moore D W, Broker T R, Chow L T . Vorinostat, a pan-HDAC inhibitor, abrogates productive HPV-18 DNA amplification. Proceedings of the National Academy of Sciences of the United States of America, 2018, 115( 47): E11138–E11147

[42] Huang Y, Chen X, Wang L, Wang T, Tang X, Su X . Centromere protein F (CENPF) serves as a potential prognostic biomarker and target for human hepatocellular carcinoma. Journal of Cancer, 2021, 12( 10): 2933–2951

[43] Shipley W U, Seiferheld W, Lukka H R, Major P P, Heney N M, Grignon D J, Sartor O, Patel M P, Bahary J P, Zietman A L, Pisansky T M, Zeitzer K L, Lawton C A F, Feng F Y, Lovett R D, Balogh A G, Souhami L, Rosenthal S A, Kerlin K J, Dignam J J, Pugh S L, Sandler H M . Radiation with or without antiandrogen therapy in recurrent prostate cancer. The New England Journal of Medicine, 2017, 376( 5): 417–428

[44] Chen Y C, Chen I S, Huang G J, Kang C H, Wang K C, Tsao M J, Pan H W . Targeting DTL induces cell cycle arrest and senescence and suppresses cell growth and colony formation through TPX2 inhibition in human hepatocellular carcinoma cells. OncoTargets and Therapy, 2018, 11: 1601–1616

[45] He X, Wang J, Zhou R, Yu S, Jiang J, Zhou Q . Kinesin family member 23 exerts a protumor function in breast cancer via stimulation of the Wnt/β-catenin pathway. Toxicology and Applied Pharmacology, 2022, 435: 115834

[46] Ma M, Lei X, Zhang Y. A review of drug-related associations prediction based on artificial intelligence methods. Current Bioinformatics, 2024, 19(6): 530−550

RIGHTS & PERMISSIONS

Higher Education Press