陆军军医大学(第三军医大学)军事预防医学系军队卫生统计学教研室,重庆
WU Yazhou, E-mail: asiawu@tmmu.edu.cn
Received:11 December 2025,
Revised:2026-01-26,
Published:30 March 2026
移动端阅览
秦茂洋,沈俊杰,王龙昊,等.MoACG:一种基于自注意力机制和门控融合的多组学与临床数据整合模型用于多癌种预后预测[J].陆军军医大学学报,2026,48(06):809-821.DOI:10.16016/j.2097-0927.202512061.
QIN M Y,SHEN J J,WANG L H,et al.MoACG: A self-attention and gated fusion-based multi-omics and clinical data integration model for pan-cancer prognosis prediction[J].J Army Med Univ,2026,48(06):809-821.DOI:10.16016/j.2097-0927.202512061.
秦茂洋,沈俊杰,王龙昊,等.MoACG:一种基于自注意力机制和门控融合的多组学与临床数据整合模型用于多癌种预后预测[J].陆军军医大学学报,2026,48(06):809-821.DOI:10.16016/j.2097-0927.202512061. DOI:
QIN M Y,SHEN J J,WANG L H,et al.MoACG: A self-attention and gated fusion-based multi-omics and clinical data integration model for pan-cancer prognosis prediction[J].J Army Med Univ,2026,48(06):809-821.DOI:10.16016/j.2097-0927.202512061. DOI:
目的
2
旨在对癌症多组学数据和临床数据进行深度融合和分析,从而提高癌症预后的预测能力。
方法
2
获取TCGA数据库卵巢癌、肝癌及结直肠癌3个癌种的组学数据(mRNA、lncRNA和miRNA数据)和临床数据;构建一种新的癌症预后预测模型MoACG(multi-omics attention clinical gating Model),其基于自注意力机制(self-attention mechanism)探讨不同组学之间的潜在关联,利用门控融合(gated fusion)自适应地融合多组学信息和临床信息(年龄、性别、治疗方式);通过与多种机器学习方法对比验证模型的有效性,最后利用可解释性算法DeepLIFT量化基因对模型贡献,筛选核心预后基因。
结果
2
在卵巢癌、肝癌及结直肠癌数据集中,五折交叉验证的AUC值分别为(0.793±0.042)、(0.791±0.065)和(0.789±0.086);AUPR值分别为(0.915±0.020)、(0.855±0.058)和(0.917±0.039);综合性能优于其他9种机器学习模型。消融实验证明,3种组学数据整合模型在各癌种中均展现出最优预测性能。DeepLIFT算法筛选出与肝癌相关的
MED8、DLGAP4、NABP2
等基因,与已有研究结论高度吻合,且能够通过表达水平有效区分患者的生存风险(
P
<
0.005)。
结论
2
相较于既往研究,利用组学数据和临床数据构建的MoACG可有效提高癌症预后预测能力,为癌症诊疗、预后研究提供新方案。
Objective
2
To perform in-depth integration and analysis of cancer multi-omics data and clinical data to enhance the predictive capability for cancer prognosis.
Methods
2
Multi-omics data (mRNA
lncRNA
and miRNA profiles) and clinical data for 3 cancer types
namely ovarian cancer
liver cancer
and colorectal cancer
were retrieved from The Cancer Genome Atlas (TCGA) database. A novel cancer prognosis prediction model
MoACG (Multi-omics Attention Clinical Gating Model)
was constructed based on the self-attention mechanism to explore potential associations among different omics layers
and gated fusion was employed to adaptively integrate multi-omics information with clinical information(age
sex
and treatment modality) . The effectiveness of the model was validated through comparison with multiple machine learning methods
and the interpretability algorithm DeepLIFT was utilized to quantify gene contributions to the model and identify core prognostic genes.
Results
2
In the ovarian cancer
liver cancer
and colorectal cancer datasets
five-fold cross-validation yielded the area under curve (AUC) values of receiver operating characteristic (ROC) curve of (0.793±0.042)
(0.791±0.065)
and (0.789±0.086)
respectively
and the AUC values of precision-recall curve (AUPR) were (0.915±0.020)
(0.855±0.058)
and (0.917±0.039)
respectively. The comprehensive performance surpassed that of 9 other machine learning models. Ablation experiments demonstrated that the 3-omics data integration model exhibited optimal predictive performance across all cancer types. The DeepLIFT algorithm identified
MED8
DLGAP4
and
NABP2
as genes associated with liver cancer
showing high concordance with existing research findings and effectively stratifying patient survival risk based on expression levels (
P
<
0.005).
Conclusion
2
Compared with previous studies
the MoACG model
constructed by integrating multi-omics and clinical data
effectively enhances the predictive performance for cancer prognosis
thereby providing a novel approach for cancer diagnosis
treatment
and prognostic research.
FORCE L M , KOCARNIK J M , MAY M L , et al . The global, regional, and national burden of cancer, 1990-2023, with forecasts to 2050: a systematic analysis for the Global Burden of Disease Study 2023 [J]. Lancet , 2010 , 406 ( 10512 ): 1565 - 1586 . DOI: 10.1016/S0140-6736(25)01635-6 http://dx.doi.org/10.1016/S0140-6736(25)01635-6 .
BRAY F , LAVERSANNE M , SUNG H , et al . Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries [J]. CA Cancer J Clin , 2024 , 74 ( 3 ): 229 - 263 . DOI: 10.3322/caac.21834 http://dx.doi.org/10.3322/caac.21834 .
TAKAGI M , KIM S , SUZUKI M , et al . Factors associated with postoperative recurrence in perforated colorectal cancer: unraveling the high recurrence rate of perforated colorectal cancer [J]. World J Surg Oncol , 2025 , 23 ( 1 ): 139 . DOI: 10.1186/s12957-025-03783-5 http://dx.doi.org/10.1186/s12957-025-03783-5 .
GIANNONE F , CASSESE G , CUBISINO A , et al . Patterns of site and timing of recurrence after curative resection in single and multiple large hepatocellular carcinoma: a multicenter international comprehensive analysis [J]. Ann Surg Oncol , 2025 , 32 ( 9 ): 6312 - 6323 . DOI: 10.1245/s10434-025-17673-3 http://dx.doi.org/10.1245/s10434-025-17673-3 .
SALAS BOLÍVAR P , GONZALEZ-BENITEZ C , CARBONELL LÓPEZ M , et al . Prognostic factors after the first recurrence of ovarian cancer [J]. J Clin Med , 2025 , 14 ( 2 ): 470 . DOI: 10.3390/jcm14020470 http://dx.doi.org/10.3390/jcm14020470 .
李鹏鹏 , 陈锡程 , 黄金羽 , 等 . 基于转录组学数据和堆叠监督自编码器的肺腺癌预后预测研究 [J]. 陆军军医大学学报 , 2023 , 45 ( 6 ): 579 - 585 . DOI: 10.16016/j.2097-0927.202212025 http://dx.doi.org/10.16016/j.2097-0927.202212025 .
LI P P , CHEN X C , HUANG J Y , et al . Prognostic prediction of lung adenocarcinoma based on transcriptomic data and stacked supervised autoencoder [J]. J Army Med Univ , 2023 , 45 ( 6 ): 579 - 585 . DOI: 10.16016/j.2097-0927.202212025 http://dx.doi.org/10.16016/j.2097-0927.202212025 .
SEOANE J , DE MATTOS-ARRUDA L . The challenge of intratumour heterogeneity in precision medicine [J]. J Intern Med , 2014 , 276 ( 1 ): 41 - 51 . DOI: 10.1111/joim.12240 http://dx.doi.org/10.1111/joim.12240 .
李佳芮 , 钱力 , 沈俊杰 , 等 . 基于转录组学和变分自编码器的癌症分期诊断研究 [J]. 陆军军医大学学报 , 2025 , 47 ( 6 ): 613 - 622 . DOI: 10.16016/j.2097-0927.202412079 http://dx.doi.org/10.16016/j.2097-0927.202412079 .
LI J R , QIAN L , SHEN J J , et al . Cancer staging diagnosis based on transcriptomics and variational autoencoder [J]. J Army Med Univ , 2025 , 47 ( 6 ): 613 - 622 . DOI: 10.16016/j.2097-0927.202412079 http://dx.doi.org/10.16016/j.2097-0927.202412079 .
QIAO T , YIN B N , LIANG J , et al . Implications of T cell-mediated tumor killing genes for molecular heterogeneity and clinical stratification in lung adenocarcinoma [J]. Genes Dis , 2024 , 11 ( 4 ): 101162 . DOI: 10.1016/j.gendis.2023.101162 http://dx.doi.org/10.1016/j.gendis.2023.101162 .
TEREKHANOVA N V , KARPOVA A , LIANG W W , et al . Epigenetic regulation during cancer transitions across 11 tumour types [J]. Nature , 2023 , 623 ( 7986 ): 432 - 441 . DOI: 10.1038/s41586-023-06682-5 http://dx.doi.org/10.1038/s41586-023-06682-5 .
BEAUDE A , AUGÉ F , ZEHRAOUI F , et al . CrossAttOmics: multiomics data integration with cross-attention [J]. Bioinformatics , 2025 , 41 ( 6 ): btaf302 . DOI: 10.1093/bioinformatics/btaf302 http://dx.doi.org/10.1093/bioinformatics/btaf302 .
SUN Q W , CHENG L , MENG A , et al . SADLN: Self-attention based deep learning network of integrating multi-omics data for cancer subtype recognition [J]. Front Genet , 2023 , 13 : 1032768 . DOI: 10.3389/fgene.2022.1032768 http://dx.doi.org/10.3389/fgene.2022.1032768 .
YE X C , SHI T Y , HUANG D , et al . Multi-Omics clustering by integrating clinical features from large language model [J]. Methods , 2025 , 239 : 64 - 71 . DOI: 10.1016/j.ymeth.2025.03.017 http://dx.doi.org/10.1016/j.ymeth.2025.03.017 .
PENG Q Q , HE Z H , LIU S C , et al . Integrating multi-omics data into a gated graph convolutional networks for identifying cancer driver genes and function modules [C]// 2023 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) . December 5 - 8 , 2023 . Istanbul, Turkiye . IEEE , 2023. DOI: 10.1109/bibm58861. 2023. 10385796.. DOI:10.1109/bibm58861.2023.10385796 http://dx.doi.org/10.1109/bibm58861.2023.10385796..DOI:10.1109/bibm58861.2023.10385796 .
YU Z G , LI Z X , LI P . MultiRepPI: a cross-modal feature fusion-based multiple characterization framework for plant peptide-protein interaction prediction [J]. BMC Plant Biol , 2025 , 25 ( 1 ): 933 . DOI: 10.1186/s12870-025-06878-z http://dx.doi.org/10.1186/s12870-025-06878-z .
XU Y , ZHANG L , SHEN X . Multi-modal adaptive gated mechanism for visual question answering [J]. PLoS One , 2023 , 18 ( 6 ): e0287557 . DOI: 10.1371/journal.pone. 0287557 http://dx.doi.org/10.1371/journal.pone.0287557 .
GONG C F , HUANG Y L , JIAN J M , et al . Measurement-guided therapeutic-dose prediction using multi-level gated modality-fusion model for volumetric-modulated arc radiotherapy [J]. Front Oncol , 2025 , 15 : 1468232 . DOI: 10.3389/fonc.2025.1468232 http://dx.doi.org/10.3389/fonc.2025.1468232 .
WANG F G , ZHUANG Z F , GAO F , et al . TMO-Net: an explainable pretrained multi-omics model for multi-task learning in oncology [J]. Genome Biol , 2024 , 25 ( 1 ): 149 . DOI: 10.1186/s13059-024-03293-9 http://dx.doi.org/10.1186/s13059-024-03293-9 .
SHRIKUMAR A , GREENSIDE P , KUNDAJE A B . Learning important features through propagating activation differences [C]// International Conference on Machine Learning ., 2017
TASAKI S , GAITERI C , MOSTAFAVI S , et al . Deep learning Decodes the principles of differential gene expression [J]. Nat Mach Intell , 2020 , 2 ( 7 ): 376 - 386 . DOI: 10.1038/s42256-020-0201-6 http://dx.doi.org/10.1038/s42256-020-0201-6 .
HAO J , KIM Y , KIM T K , et al . PASNet: pathway-associated sparse deep neural network for prognosis prediction from high-throughput data [J]. BMC Bioinformatics , 2018 , 19 ( 1 ): 510 . DOI: 10.1186/s12859-018-2500-z http://dx.doi.org/10.1186/s12859-018-2500-z .
YAN H X , WENG D W , LI D G , et al . Prior knowledge-guided multilevel graph neural network for tumor risk prediction and interpretation via multi-omics data integration [J]. Brief Bioinform , 2024 , 25 ( 3 ): bbae184 . DOI: 10.1093/bib/bbae184 http://dx.doi.org/10.1093/bib/bbae184 .
ZHANG G , PENG Z , YAN C K , et al . MultiGATAE: a novel cancer subtype identification method based on multi-omics and attention mechanism [J]. Front Genet , 2022 , 13 : 855629 . DOI: 10.3389/fgene.2022.855629 http://dx.doi.org/10.3389/fgene.2022.855629 .
ZHAO C M , GAO F , SHEN Z H . Multi-motion sensor behavior based continuous authentication on smartphones using gated two-tower transformer fusion networks [J]. Comput Secur , 2024 , 139 : 103698 . DOI: 10.1016/j.cose.2023.103698 http://dx.doi.org/10.1016/j.cose.2023.103698 .
ELMARAKEBY H A , HWANG J , ARAFEH R , et al . Biologically informed deep neural network for prostate cancer discovery [J]. Nature , 2021 , 598 ( 7880 ): 348 - 352 . DOI: 10.1038/s41586-021-03922-4 http://dx.doi.org/10.1038/s41586-021-03922-4 .
LAN W , LIAO H B , CHEN Q F , et al . DeepKEGG: a multi-omics data integration framework with biological insights for cancer recurrence prediction and biomarker discovery [J]. Brief Bioinform , 2024 , 25 ( 3 ): bbae185 . DOI: 10.1093/bib/bbae185 http://dx.doi.org/10.1093/bib/bbae185 .
SUN R , JUN A , YU H L , et al . Proteomic landscape profiling of primary prostate cancer reveals a 16-protein panel for prognosis prediction [J]. Cell Rep Med , 2024 , 5 ( 8 ): 101679 . DOI: 10.1016/j.xcrm.2024.101679 http://dx.doi.org/10.1016/j.xcrm.2024.101679 .
ZHENG Y Y , WANG G H , JIANG A . Explore the expression of mitochondria-related genes to construct prognostic risk model for ovarian cancer and validate it, so as to provide optimized treatment for ovarian cancer [J]. Front Immunol , 2024 , 15 : 1458264 . DOI: 10.3389/fimmu.2024.1458264 http://dx.doi.org/10.3389/fimmu.2024.1458264 .
WU Y , XU Y J , CHEN Y T , et al . Development of a prognostic prediction model based on damage-associated molecular pattern for colorectal cancer applying bulk RNA-seq analysis [J]. Sci Rep , 2025 , 15 ( 1 ): 25792 . DOI: 10.1038/s41598-025-10592-z http://dx.doi.org/10.1038/s41598-025-10592-z .
CUI H T , WANG C , MAAN H , et al . scGPT: toward building a foundation model for single-cell multi-omics using generative AI [J]. Nat Methods , 2024 , 21 ( 8 ): 1470 - 1480 . DOI: 10.1038/s41592-024-02201-0 http://dx.doi.org/10.1038/s41592-024-02201-0 .
DONG B T , ZHANG H , DUAN Y Y , et al . Development of a machine learning-based model to predict prognosis of alpha-fetoprotein-positive hepatocellular carcinoma [J]. J Transl Med , 2024 , 22 ( 1 ): 455 . DOI: 10.1186/s12967-024-05203-w http://dx.doi.org/10.1186/s12967-024-05203-w .
MAO R J , WAN L , ZHOU M H , et al . Cox-Sage: enhancing Cox proportional hazards model with interpretable graph neural networks for cancer prognosis [J]. Brief Bioinform , 2025 , 26 ( 2 ): bbaf108 . DOI: 10.1093/bib/bbaf108 http://dx.doi.org/10.1093/bib/bbaf108 .
JOSE A , SRIVASTAVA A , ANSARI A , et al . DeepGraphMut: a graph-based deep learning method for cancer prognosis using somatic mutation profile [J]. Brief Bioinform , 2025 , 26 ( 4 ): bbaf409 . DOI: 10.1093/bib/bbaf409 http://dx.doi.org/10.1093/bib/bbaf409 .
YING Y , WU N , HUO J H , et al . scUCAF: an uncertainty-aware cross-omics alignment and fusion network for single-cell multi-omics data clustering [J]. Comput Biol Chem , 2026 , 120 : 108631 . DOI: 10.1016/j.compbiolchem. 2025. 108631 http://dx.doi.org/10.1016/j.compbiolchem.2025.108631 .
WANG J H , LIAO N Q , DU X F , et al . A semi-supervised approach for the integration of multi-omics data based on transformer multi-head self-attention mechanism and graph convolutional networks [J]. BMC Genomics , 2024 , 25 ( 1 ): 86 . DOI: 10.1186/s12864-024-09985-7 http://dx.doi.org/10.1186/s12864-024-09985-7 .
CHI G , PEI J H , LI X Q . RAE1 is a prognostic biomarker and is correlated with clinicopathological characteristics of patients with hepatocellular carcinoma [J]. BMC Bioinformatics , 2022 , 23 ( 1 ): 252 . DOI: 10.1186/s12859-022-04806-8 http://dx.doi.org/10.1186/s12859-022-04806-8 .
NI Y , LU M Q , LI M , et al . Unraveling the underlying pathogenic factors driving nonalcoholic steatohepatitis and hepatocellular carcinoma: an in-depth analysis of prognostically relevant gene signatures in hepatocellular carcinoma [J]. J Transl Med , 2024 , 22 ( 1 ): 72 . DOI: 10.1186/s12967-024-04885-6 http://dx.doi.org/10.1186/s12967-024-04885-6 .
DONG C R , HUANG S L , SUN L , et al . DLGAP4 acts as an effective prognostic predictor for hepatocellular carcinoma and is closely related to tumour progression [J]. Sci Rep , 2022 , 12 ( 1 ): 19775 . DOI: 10.1038/s41598-022-23837-y http://dx.doi.org/10.1038/s41598-022-23837-y .
DAI C C , TONG Y , BAI N , et al . Decoding the role of nucleic acid binding protein 2 in lipid dysregulation and hepatocellular carcinoma progression through LKB1-mediated mitochondrial dysfunction [J]. Cell Signal , 2025 , 132 : 111820 . DOI: 10.1016/j.cellsig.2025.111820 http://dx.doi.org/10.1016/j.cellsig.2025.111820 .
WILLS E S , CNOSSEN W R , VELTMAN J A , et al . Chromosomal abnormalities in hepatic cysts point to novel polycystic liver disease genes [J]. Eur J Hum Genet , 2016 , 24 ( 12 ): 1707 - 1714 . DOI: 10.1038/ejhg.2016.97 http://dx.doi.org/10.1038/ejhg.2016.97 .
DAHABIEH M S , HUANG F , GONCALVES C , et al . Silencing PEX26 as an unconventional mode to kill drug-resistant cancer cells and forestall drug resistance [J]. Autophagy , 2022 , 18 ( 3 ): 540 - 558 . DOI: 10.1080/15548627.2021.1936932 http://dx.doi.org/10.1080/15548627.2021.1936932 .
ZHOU P H , SHAFFER D R , ALVAREZ ARIAS D A , et al . In vivo discovery of immunotherapy targets in the tumour microenvironment [J]. Nature , 2014 , 506 ( 7486 ): 52 - 57 . DOI: 10.1038/nature12988 http://dx.doi.org/10.1038/nature12988 .
SONG J , ZHENG J , LIU X B , et al . A novel protein encoded by ZCRB1-induced circHEATR5B suppresses aerobic glycolysis of GBM through phosphorylation of JMJD5 [J]. J Exp Clin Cancer Res , 2022 , 41 ( 1 ): 171 . DOI: 10.1186/s13046-022-02374-6 http://dx.doi.org/10.1186/s13046-022-02374-6 .
HAMADA K , NAKANISHI Y , YU M T , et al . Retinoblastoma-binding protein 9 suppresses intestinal inflammation and inflammation-induced tumorigenesis in mice [J]. Cell Mol Gastroenterol Hepatol , 2025 , 19 ( 3 ): 101435 . DOI: 10.1016/j.jcmgh.2024.101435 http://dx.doi.org/10.1016/j.jcmgh.2024.101435 .
0
Views
5
下载量
0
CSCD
Publicity Resources
Related Articles
Related Author
Related Institution
京公网安备11010802024621
