微重力生物学知识图谱三元关系预测模型构建及应用
doi: 10.11728/cjss2025.02.2024-0167 cstr: 32142.14.cjss.2024-0167
-
朱学松 男, 2004年5月出生于山西省大同市, 现就读于华中农业大学信息学院生物信息系, 主要研究方向为空间疾病致病机制解析、药物发现等. E-mail: 1474452250@qq.com -
曲恩锐 男, 2004年11月出生于内蒙古自治区包头市, 现就读于华中农业大学信息学院生物信息系, 主要研究方向为空间疾病致病机制解析、药物发现、数学建模等. E-mail: 3315690090@qq.com -
朱玉锋 男, 2001年11月出生于江苏省泰州市, 毕业于华中农业大学信息学院生物信息系, 主要研究方向为空间生物医学、药物筛选等. E-mail: neverbo1101@outlook.com -
全源 女, 1990年7月出生于广西壮族自治区柳州市, 现为华中农业大学信息学院副研究员, 博士生导师, 主要研究方向为空间生物信息学、空间医学、药物发现等. E-mail: quanyuan@mail.hzau.edu.cn
作者简介:
通讯作者:
Construction and Application of a Ternary Relationship Prediction Model for Microgravity Biological Knowledge Graph
-
摘要: 空间微重力环境会对航天员的生理和心理健康造成重大影响, 包括骨骼密度下降、肌肉萎缩、心血管功能改变等. 这些问题对于实现长期空间居住和深空探测构成重大障碍. 为应对这些问题, 研究通过整合微重力生物学知识图谱(Microgravity Biomedical Knowledge Graph, MBKG)和药物重定位知识图谱(Drug Repurposing Knowledge Graph, DRKG), 构建一个全面的知识图谱, 覆盖广泛的疾病、药物和基因实体及实体之间的复杂关系. 在此基础上, 研究训练并使用新的三元关系预测模型异质因果元路径图神经网络(Heterogeneous Causal Meta path Graph Neural Network, HCMGNN)获取预测结果. 结果表明, 与传统的知识图谱二元链路预测相比, 研究提出的三元预测方法在提高基因和药物预测准确率方面具有明显优势. 研究结论强调三元关系模型在探测基因–药物–疾病三元关系预测方面的有效性和潜力, 不仅为未来空间探测中航天员的生理和心理健康、药物再利用研究提供了新的方法和研究思路, 而且为药物再利用领域开辟了新的视角.Abstract: With the advancement of science and technology, the demand for space exploration has become particularly urgent. However, the microgravity environment in space has negative impacts on the physiological and psychological health of astronauts, including decreased bone density, muscle atrophy, and changes in cardiovascular function. These challenges pose significant barriers to the realization of long-term space habitation and deep space exploration. To address these challenges, this study integrates Microgravity Biomedical Knowledge Graphs (MBKG) and Drug Repurposing Knowledge Graphs (DRKG) to construct a comprehensive knowledge graph that covers a wide range of diseases, drugs, and genes, as well as the complex relationships between entities. Based on this, the study trains and uses a new ternary relationship prediction model, Heterogeneous Causal Meta path Graph Neural Network (HCMGNN), to obtain prediction results. The results show that compared with traditional binary link prediction in knowledge graphs, the ternary prediction method proposed in this study has a significant advantage in improving the accuracy of gene and drug predictions. The study concludes that the ternary relationship model is effective and has the potential to explore the prediction of gene-drug-disease ternary relationships, provide new methods and research ideas for the physiological and psychological health of astronauts in future space exploration and drug repurposing research, and opening up new perspectives in the field of drug repurposing.
-
图 3 五折交叉验证和独立性验证EHITS-N与Average MRR结果折线. (a)独立性验证EHITS-N结果折线, (b)～(f)五折交叉验证EHITS-N结果折线. (g)独立性验证Average MRR结果折线, (h)～(l)五折交叉验证Average MRR结果折线
Figure 3. Line graphs of EHITS-N and Average MRR Results for 5-fold cross-validation and independence validation. (a) Independence validation EHITS-N result line graph, (b)～(f) five-fold cross-validation EHITS-N result line graphs. (g) independence validation Average MRR result line graph, (h)～(l) five-fold cross-validation Average MRR result line graphs
图 4 五折交叉验证和独立性验证Average NDCG结果折线. (a)独立性验证Average NDCG结果折线, (b)～(f)为五折交叉验证Average NDCG结果折线
Figure 4. Line graphs of Average NDCG results for 5-fold cross-validation and independence validation. (a) Independence validation Average NDCG result line graphs, (b)～(f) the five-fold cross-validation Average NDCG result line graphs. The horizontal scale shows the change in Epcho, the vertical scale shows the change in Average NDCG
表 1 实体格式统一示例
Table 1. Examples of entities formatting uniformity
数据原格式 统一后格式 ENTREZ:174 Gene::174 MESH:C000123 Compound::MESH::C000123 MESH:D011471 Disease::MESH::D011471 
下载: 导出CSV
表 2 部分MBKG关系补全
Table 2. Alignment of MBKG relations
MBKG关系 对齐 缩写 chemical:gene:Z GNBR::Z::Compound:Gene Z gene:disease:Y GNBR::Y::Gene:Disease Y gene:disease:X GNBR::X::Gene:Disease X gene:gene:W GNBR::W::Gene:Gene W gene:gene:V+ GNBR::V+::Gene:Gene V+ gene:disease:Ud GNBR::Ud::Gene:Disease Ud gene:disease:U GNBR::U::Gene:Disease U gene:disease:Te GNBR::Te::Gene:Disease Te chemical:disease:T GNBR::T::Compound:Disease T chemical:disease:Sa GNBR::Sa::Compound:Disease Sa gene:gene:Rg GNBR::Rg::Gene:Gene Rg gene:gene:Q GNBR::Q::Gene:Gene Q chemical:disease:Pr GNBR::Pr::Compound:Disease Pr chemical:disease:Pa GNBR::Pa::Compound:Disease Pa chemical:gene:O GNBR::O::Compound:Gene O … … … … … …
下载: 导出CSV
表 3 TransE模型嵌入训练结果
Table 3. Results of TransE model embedding training
评价指标 结果 Average MRR 0.737 Average MR 6.67 Average HITS-10 0.640 Average HITS-3 0.804 Average HITS-10 0.911
下载: 导出CSV
表 4 五折交叉验证结果与独立性验证结果
Table 4. Results of 5-fold cross-validation and the results of independence validation
五折交叉验证结果 独立性验证结果 评价指标 结果 评价指标 结果 Average HITS-1 0.761 Average HITS-1 0.865 Average HITS-3 0.944 Average HITS-3 0.980 Average HITS-5 0.981 Average HITS-5 0.990 Average NDCG-1 0.761 Average NDCG-1 0.865 Average NDCG-3 0.872 Average NDCG-3 0.872 Average NDCG-5 0.886 Average NDCG-5 0.934 Average MRR 0.857 Average MRR 0.921
下载: 导出CSV
表 5 预测得分前10的基因–药物–疾病三元关系
Table 5. Gene-compound-disease ternary relationships with Top 10 predicted scores
Rank Gene_id Gene_name Compound_id Compound_name Score 1 Gene::632 BGLAP Compound::DB01645 Genistein –1.96359 2 Gene::1588 CYP19 A1 Compound::DB01645 Genistein –2.00393 3 Gene::5468 PPARγ Compound::DB01645 Genistein –2.06026 4 Gene::7040 TGF-β1 Compound::DB01645 Genistein –2.06260 5 Gene::2099 ESR1 Compound::DB06202 Lasofoxifene –2.13473 6 Gene::5745 PTH1 R Compound::DB06285 Teriparatide –2.16564 7 Gene::2908 NR3 C1 Compound::DB00624 Testosterone –2.19646 8 Gene::268 AMH Compound::DB01234 Dexamethasone –2.41574 9 Gene::2099 ESR1 Compound::DB06249 Arzoxifene –2.42039 10 Gene::5745 PTH1 R Compound::DB01234 Dexamethasone –2.43293
下载: 导出CSV
-
[1] KIM D H, LIN S C. Human capital and natural resource dependence[J]. Structural Change and Economic Dynamics, 2017, 40: 92-102 doi: 10.1016/j.strueco.2017.01.002 [2] DREES J M, HEUGENS P P M A R. Synthesizing and extending resource dependence theory: a meta-analysis[J]. Journal of Management, 2013, 39(6): 1666-1698 doi: 10.1177/0149206312471391 [3] FANKHAUSER T, WANG Q, GERLICHER A, et al. Resource dependency processing in web scaling frameworks[J]. IEEE Transactions on Services Computing, 2018, 11(1): 155-168 doi: 10.1109/TSC.2016.2561934 [4] DICKERSON B L, SOWINSKI R, KREIDER R B, et al. Impacts of microgravity on amino acid metabolism during spaceflight[J]. Experimental Biology and Medicine, 2023, 248(5): 380-393 doi: 10.1177/15353702221139189 [5] OLUWAFEMI F A, ABDELBAKI R, LAI J C Y, et al. A review of astronaut mental health in manned missions: Potential interventions for cognitive and mental health challenges[J]. Life Sciences in Space Research, 2021, 28: 26-31 doi: 10.1016/j.lssr.2020.12.002 [6] SPRUGNOLI G, CAGLE Y D, SANTARNECCHI E. Microgravity and cosmic radiations during space exploration as a window into neurodegeneration on earth[J]. JAMA Neurology, 2020, 77(2): 157-158 doi: 10.1001/jamaneurol.2019.4003 [7] VAN OMBERGEN A, DEMERTZI A, TOMILOVSKAYA E, et al. The effect of spaceflight and microgravity on the human brain[J]. Journal of Neurology, 2017, 264(1): 18-22 [8] ZHENG D, SONG X, MA C, et al. DGL-KE: training knowledge graph embeddings at scale[C]//Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York: ACM, 2020 [9] SANG S T, YANG Z H, LIU X X, et al. GrEDeL: a knowledge graph embedding based method for drug discovery from biomedical literatures[J]. IEEE Access, 2018, 7: 8404-8415 [10] FARRUGIA L, AZZOPARDI L M, DEBATTISTA J, et al. Predicting drug-drug interactions using knowledge graphs[OL]. arXiv preprint arXiv: 2308.04172, 2023 [11] ZHENG Y H, PAN G J, QUAN Y, et al. Construction of microgravity biological knowledge graph and its applications in anti-osteoporosis drug prediction, Life Sciences in Space Research, Volume 41, 2024, Pages 64-73, ISSN 2214-5524. [12] Ioannidis, V. N. et al. Drkg-drug repurposing knowledge graph for covid-19. GitHub https://github.com/gnn4dr /DRKG (2020). Accessed 01 Jan 2022. [13] ZHU G H, ZHU Z N, CHEN H Y, et al. HAGNN: hybrid aggregation for heterogeneous graph neural networks[OL]. arXiv preprint arXiv: 2307.01636, 2023 [14] LIANG T, LIU J. Meta-path generation online for heterogeneous network embedding[C]//Proceedings of the 2020 International Joint Conference on Neural Networks (IJCNN). Glasgow, UK: IEEE, 2020 [15] RODRIGUEZ J D, PEREZ A, LOZANO J A. Sensitivity analysis of k-fold cross validation in prediction error estimation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(3): 569-575 doi: 10.1109/TPAMI.2009.187 [16] POHJANKUKKA J, PAHIKKALA T, NEVALAINEN P, et al. Estimating the prediction performance of spatial models via spatial k-fold cross validation[J]. International Journal of Geographical Information Science, 2017, 31(10): 2001-2019 doi: 10.1080/13658816.2017.1346255 [17] BERGMEIR C, HYNDMAN R J, KOO B. A note on the validity of cross-validation for evaluating autoregressive time series prediction[J]. Computational Statistics :Times New Roman;">& Data Analysis, 2018, 120: 70-83 [18] SAHA S, ROY D, MITRA M. On modifying evaluation measures to deal with ties in ranked lists[C]//Proceedings of the 22nd ACM/IEEE Joint Conference on Digital Libraries. Cologne, Germany: IEEE, 2022 [19] GOEL S, KUMAR R, KUMAR M, et al. An efficient page ranking approach based on vector norms using sNorm(p) algorithm[J]. Information Processing :Times New Roman;">& Management, 2019, 56(3): 1053-1066 [20] SADEGHI A, GRAUX D, YAZDI H S, et al. MDE: multiple distance embeddings for link prediction in knowledge graphs[OL]. arXiv preprint arXiv: 1905.10702, 2019 [21] TURAL S, KARA N, ALAYLI G, et al. Association between osteoporosis and polymorphisms of the bone Gla protein, estrogen receptor 1, collagen 1-A1 and calcitonin receptor genes in Turkish postmenopausal women[J]. Gene, 2013, 515(1): 167-172 doi: 10.1016/j.gene.2012.10.041 [22] SEHMISCH S, UFFENORDE J, MAEHLMEYER S, et al. Evaluation of bone quality and quantity in osteoporotic mice–the effects of genistein and equol[J]. Phytomedicine, 2010, 17(6): 424-430 doi: 10.1016/j.phymed.2009.10.004 [23] Da Silva F R P, Vasconcelos A C C G, Casimiro G S, et al. Quantitative assessment of the association between polymorphisms in osteoprotegerin gene and risk of low bone mineral density[J]. International Archives of Medicine, 2015: 8(169) [24] ŞIRIN Ö K, AYDOĞAN H Y, UYAR M, et al. PPARγ Pro12Ala and C161T polymorphisms, but not PPARα L162V, are associated with osteoporosis risk in Turkish postmenopausal women[J]. İstanbul Journal of Pharmacy, 2019, 49(1): 14-19 [25] SCHOLTYSEK C, KATZENBEISSER J, FU H, et al. PPARβ/δ governs Wnt signaling and bone turnover[J]. Nature Medicine, 2013, 19(5): 608-613 doi: 10.1038/nm.3146 [26] ZAYZAFOON M, GATHINGS W E, MCDONALD J M. Modeled microgravity inhibits osteogenic differentiation of human mesenchymal stem cells and increases Adipogenesis[J]. Endocrinology, 2004, 145(5): 2421-2432 doi: 10.1210/en.2003-1156 [27] MANN V, GRIMM D, CORYDON T J, et al. Changes in human foetal osteoblasts exposed to the random positioning machine and bone construct tissue engineering[J]. International Journal of Molecular Sciences, 2019, 20(6): 1357 doi: 10.3390/ijms20061357 [28] JURZAK M, ADAMCZYK K, ANTOŃCZAK P, et al. Evaluation of genistein ability to modulate CTGF mRNA/protein expression, genes expression of TGFβ isoforms and expression of selected genes regulating cell cycle in keloid fibroblasts in vitro[J]. Acta Poloniae Pharmaceutica, 2014, 71(6): 972-986 [29] Plourde P V, SCHWARTZBERG L S, GREENE G L, et al. An open-label, randomized, multi-center phase 2 study evaluating the activity of lasofoxifene relative to fulvestrant for the treatment of postmenopausal women with locally advanced or Metastatic ER+/HER2- Breast Cancer (MBC) with an ESR1 mutation[C]//Cancer Research, San Antonio: San Antonio Breast Cancer Symposium, 2019 [30] GOETZ M P, GAL-YAM E, STOVER D, et al. Abstract P5-05-04: Estrogen Receptor 1 (ESR1) mutations in circulating tumor DNA (ctDNA) from patients with ER+/HER2-metastatic Breast Cancer (mBC) treated with lasofoxifene or fulvestrant in the ELAINE 1 study[J]. Cancer Research, 2023, 83(S5): P5-05-04 [31] SEFRIOUI D, PERDRIX A, SARAFAN-VASSEUR N, et al. Short report: monitoring ESR1 mutations by circulating tumor DNA in aromatase inhibitor resistant metastatic breast cancer[J]. International Journal of Cancer, 2015, 137(10): 2513-2519 doi: 10.1002/ijc.29612 [32] NOORDIN S, GLOWACKI J. Parathyroid hormone and its receptor gene polymorphisms: implications in osteoporosis and in fracture healing[J]. Rheumatology International, 2016, 36(1): 1-6 doi: 10.1007/s00296-015-3319-9 [33] STYRKARSDOTTIR U, THORLEIFSSON G, GUDJONSSON S A, et al. Sequence variants in the PTCH1 gene associate with spine bone mineral density and osteoporotic fractures[J]. Nature Communications, 2016, 7(1): 10129 doi: 10.1038/ncomms10129 [34] NAKAMURA T, SUGIMOTO T, NAKANO T, et al. Randomized Teriparatide [Human Parathyroid Hormone (PTH) 1–34] Once-Weekly Efficacy Research (TOWER) trial for examining the reduction in new vertebral fractures in subjects with primary osteoporosis and high fracture risk[J]. The Journal of Clinical Endocrinology :Times New Roman;">& Metabolism, 2012, 97(9): 3097-3106 [35] CUMMINGS S R, MCCLUNG M, REGINSTER J Y, et al. Arzoxifene for prevention of fractures and invasive breast cancer in postmenopausal women[J]. Journal of Bone and Mineral Research, 2011, 26(2): 397-404 doi: 10.1002/jbmr.191 [36] VICO L, HARGENS A. Skeletal changes during and after spaceflight[J]. Nature Reviews Rheumatology, 2018, 14(2): 229-245 [37] SMITH S M, HEER M A, SHACKELFORD L C, et al. Benefits for bone from resistance exercise and nutrition in long-duration spaceflight[J]. Journal of Applied Physiology, 2012, 112(1): 105-113 [38] CARTER J A, BUCKEY J C, GREENHALGH L, et al. An interactive media program for managing psychosocial problems on long-duration spaceflights[J]. Aviation, Space, and Environmental Medicine, 2005, 76(S6): B213-B223 -
-
下载:
图(4) / 表(5)
计量
- 文章访问数: 271
- HTML全文浏览量: 61
- PDF下载量: 21
-
被引次数:
0(来源:Crossref)
0(来源:其他)
