Dual-BINN: a dual-branch biologically informed neural network integrating Reactome pathways and ClassyFire structures for metabolomics

Abstract

Aim: To address the limitations of current metabolomics analysis, including the neglect of metabolite interdependencies, poor interpretability of black-box models, and incomplete utilization of biological information due to pathway annotation limitations. This study aims to develop and validate a dual-branch biologically informed neural network (Dual-BINN) that integrates metabolic pathway hierarchy and molecular structural hierarchy for improved prediction and interpretability in metabolomics.

Methods: We developed a Dual-BINN, which explicitly incorporates metabolic pathway hierarchy and molecular structural category hierarchy into the model architecture. Pathway and structure subnetworks were constructed and integrated via an adaptive fusion mechanism. SHAP was employed for interpretability analysis. The model was evaluated using multi-center plasma metabolomics data for gastric cancer and a breast cancer dataset.

Results: On the independent gastric cancer test set, Dual-BINN achieved a recall of 0.937, which compares favorably with the recall of 0.905 reported by the 10-DM model on the same dataset split. Key metabolites were enriched in the tricarboxylic acid cycle, one-carbon metabolism, and energy metabolism pathways, while structurally concentrated in organic acids, amino acids, and nucleoside-related compounds. The model also demonstrated excellent classification performance on the breast cancer dataset, confirming strong cross-disease generalization ability.

Conclusions: The proposed framework enhances predictive performance while providing biologically meaningful structured interpretations, offering a robust computational approach for metabolomic mechanism analysis and biomarker discovery in complex diseases.

References

[1] Muthubharathi BC, Gowripriya T, Balamurugan K. Metabolomics: small molecules that matter more. Mol Omics. 2021; 17: 210-29.

[2] Yang Q, Cai Y, Guan Y, Wang Z, Guo S, Qiu S, et al. Metabolic phenotypes: Molecular bridges between health homeostasis and disease imbalance. Comput Struct Biotechnol J. 2025; 27: 4710-9.

[3] Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: Cancer J Clin. 2024; 74: 229-63.

[4] Han B, Zheng R, Zeng H, Wang S, Sun K, Chen R, et al. Cancer incidence and mortality in China, 2022. J Natl Cancer Cent. 2024; 4: 47-53.

[5] Fuller H, Zhu Y, Nicholas J, Chatelaine HA, Drzymalla EM, Sarvestani AK, et al. Metabolomic epidemiology offers insights into disease aetiology. Nat Metab. 2023; 5: 1656-72.

[6] Al—Sulaiti H, Almaliti J, Naman CB, Al Thani AA, Yassine HM. Metabolomics Approaches for the Diagnosis, Treatment, and Better Disease Management of Viral Infections. Metabolites. 2023; 13: 948.

[7] Sillé F, Hartung T. Metabolomics in Preclinical Drug Safety Assessment: Current Status and Future Trends. Metabolites. 2024; 14: 98.

[8] Gao X, Yan M, Zhang C, Wu G, Shang J, Zhang C, et al. MDNN—DTA: a multimodal deep neural network for drug—target affinity prediction. Front Genet. 2025; 16: 1527300.

[9] Hussain S, Xi X, Ullah I, Inam SA, Naz F, Shaheed K, et al. A Discriminative Level Set Method with Deep Supervision for Breast Tumor Segmentation. Comput Biol Med. 2022; 149: 105995.

[10] Inam SA, Iqbal D, Hashim H, Khuhro MA. An empirical approach towards detection of tuberculosis using deep convolutional neural network. Int J Data Min Model Manag. 2024; 16: 101—12.

[11] Galal A, Talal M, Moustafa A. Applications of machine learning in metabolomics: Disease modeling and classification. Front Genet. 2022; 13: 1017340.

[12] Chi J, Shu J, Li M, Mudappathi R, Jin Y, Lewis F, et al. Artificial intelligence in metabolomics: a current review. TrAC Trends Anal Chem. 2024; 178: 117852.

[13] Sen P, Lamichhane S, Mathema VB, McGlinchey A, Dickens AM, Khoomrung S, et al. Deep learning meets metabolomics: a methodological perspective. Brief Bioinform. 2020; 22: 1531—42.

[14] Elguoshy A, Zedan H, Saito S. Machine Learning—Driven Insights in Cancer Metabolomics: From Subtyping to Biomarker Discovery and Prognostic Modeling. Metabolites. 2025; 15: 514.

[15] Souto—Carneiro M, Tóth L, Behnisch R, Urbach K, Klika KD, Carvalho RA, et al. Differences in the serum metabolome and lipidome identify potential biomarkers for seronegative rheumatoid arthritis versus psoriatic arthritis. Ann Rheum Dis. 2020; 79: 499-506.

[16] Rahujo A, Atif D, Inam SA, Khan AA, Ullah S. A survey on the applications of transfer learning to enhance the performance of large language models in healthcare systems. Discov Artif Intell. 2025; 5: 90.

[17] Elmarakeby HA, Hwang J, Arafeh R, Crowdis J, Gang S, Liu D, et al. Biologically informed deep neural network for prostate cancer discovery. Nature. 2021; 598: 348—52.

[18] Hartman E, Scott AM, Karlsson C, Mohanty T, Vaara ST, Linder A, et al. Interpreting biologically informed neural networks for enhanced proteomic biomarker discovery and pathway analysis. Nat Commun. 2023; 14: 5359.

[19] Kaynar G, Cakmakci D, Bund C, Todeschi J, Namer IJ, Cicek AE. PiDeeL: metabolic pathway—informed deep learning model for survival analysis and pathological classification of gliomas. Bioinformatics. 2023; 39: e39.

[20] Chen Y, Wang B, Zhao Y, Shao X, Wang M, Ma F, et al. Metabolomic machine learning predictor for diagnosis and prognosis of gastric cancer. Nat Commun. 2024; 15: 1657.

[21] Milacic M, Beavers D, Conley P, Gong C, Gillespie M, Griss J, et al. The Reactome Pathway Knowledgebase 2024. Nucleic Acids Res. 2023; 52: D672-8.

[22] Djoumbou Feunang Y, Eisner R, Knox C, Chepelev L, Hastings J, Owen G, et al. ClassyFire: automated chemical classification with a comprehensive, computable taxonomy. J Cheminformatics. 2016; 8: 61.

[23] Lundberg SM, Lee SI. A unified approach to interpreting model predictions. In: Proceedings of the 31st International Conference on Neural Information Processing Systems; 2017 Dec 4—9; Long Beach, California, USA. Curran Associates Inc.; 2017. pp. 4768—77.

[24] Zhao Y, Li JS, Guo MZ, Feng BS, Zhang JP. Inhibitory effect of S—adenosylmethionine on the growth of human gastric cancer cells in vivo and in vitro. Chin J Cancer. 2010; 29: 752-60.

[25] Gao P, Zuo C, Yuan W, Cai J, Chai X, Gong R, et al. Spatiotemporal multi—omics analysis uncovers NAD—dependent immunosuppressive niche triggering early gastric cancer. Signal Transduct Target Ther. 2025; 10: 313.

[26] Yaku K, Okabe K, Hikosaka K, Nakagawa T. NAD Metabolism in Cancer Therapeutics. Front Oncol. 2018; 8: 622.

[27] Mu X, Zhao T, Xu C, Shi W, Geng B, Shen J, et al. Oncometabolite succinate promotes angiogenesis by upregulating VEGF expression through GPR91—mediated STAT3 and ERK activation. Oncotarget. 2017; 8: 13174—85.

[28] Tsuchiya A, Nishizaki T. Anticancer effect of adenosine on gastric cancer via diverse signaling pathways. World J Gastroenterol. 2015; 21: 10931—5.

[29] Saitoh M, Nagai K, Nakagawa K, Yamamura T, Yamamoto S, Nishizaki T. Adenosine induces apoptosis in the human gastric cancer cells via an intrinsic pathway relevant to activation of AMP—activated protein kinase. Biochem Pharmacol. 2004; 67: 2005-11.

[30] Lu Y, Zhang X, Zhang H, Lan J, Huang G, Varin E, et al. Citrate induces apoptotic cell death: a promising way to treat gastric carcinoma? Anticancer Res. 2011; 31: 797-805.

[31] Yuan LW, Yamashita H, Seto Y. Glucose metabolism in gastric cancer: The cutting—edge. World J Gastroenterol. 2016; 22: 2046-59.

[32] Wu J, Liu N, Chen J, Tao Q, Li Q, Li J, et al. The Tricarboxylic Acid Cycle Metabolites for Cancer: Friend or Enemy. Research. 2024; 7: 0351.

[33] Ahuja S, Zaheer S. Molecular Mediators of Metabolic Reprogramming in Cancer: Mechanisms, Regulatory Networks, and Therapeutic Strategies. Immunology. 2025; 177: 1-43.

[34] Liu K, Liu Y, Zhang S, Li Z, Qu W, Li P, et al. Epigenetic regulation of RNA methylations in gastric cancer. Oncol Rev. 2025; 19: 1601511.

[35] Mentch SJ, Locasale JW. One—carbon metabolism and epigenetics: understanding the specificity. Ann N Y Acad Sci. 2015; 1363: 91-8.

[36] Cao X, Guo Y, Guo Z, Liu Y, Dou Y, Xue L. One—carbon metabolism in cancer: moonlighting functions of metabolic enzymes and anti—tumor therapy. Cancer Metastasis Rev. 2025; 44: 91.

[37] Zhang M, Saad C, Le L, Halfter K, Bauer B, Mansmann UR, et al. Computational modeling of methionine cycle—based metabolism and DNA methylation and the implications for anti—cancer drug response prediction. Oncotarget. 2018; 9: 22546—58.

[38] Borin TF, Angara K, Rashid MH, Achyut BR, Arbab AS. Arachidonic Acid Metabolite as a Novel Therapeutic Target in Breast Cancer Metastasis. Int J Mol Sci. 2017; 18: 2661.

[39] Ghanbari F, Fortier AM, Park M, Philip A. Cholesterol—Induced Metabolic Reprogramming in Breast Cancer Cells Is Mediated via the ERRα Pathway. Cancers. 2021; 13: 2605.

[40] Wang H, Sun J, Jiang Z, Tong Z, Wang C. Glutamate promotes triple—negative breast cancer development through IRE1α/XBP1—mediated macrophage polarization: mechanism insights and therapy. Discov Oncol. 2025; 16: 1009.

[41] Knott SRV, Wagenblast E, Khan S, Kim SY, Soto M, Wagner M, et al. Asparagine bioavailability governs metastasis in a model of breast cancer. Nature. 2018; 554: 378-81.

[42] Jiang Y, Cao Y, Wang Y, Li W, Liu X, Lv Y, et al. Cysteine transporter SLC3A1 promotes breast cancer tumorigenesis. Theranostics. 2017; 7: 1036-46.

[43] Lu L, Xue L, Jiang S, He G, Deng X. The pseudouridine synthase PUS7 is associated with stemness and represents a potential therapeutic target in triple—negative breast cancer cells. Sci Rep. 2026; 16: 2411.

[44] Xie G, Zhou B, Zhao A, Qiu Y, Zhao X, Garmire L, et al. Lowered circulating aspartate is a metabolic feature of human breast cancer. Oncotarget. 2015; 6: 33369—81.