图自动编码器上二阶段融合实现的环状RNA-疾病关联预测

doi:10.11772/j.issn.1001-9081.2022050727

《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (6): 1979-1986.DOI: 10.11772/j.issn.1001-9081.2022050727

• 前沿与综合应用 • 上一篇

图自动编码器上二阶段融合实现的环状RNA-疾病关联预测

张奕¹^,², 王真梅¹()

^1.桂林理工大学信息科学与工程学院, 广西桂林 541006
^2.广西嵌入式技术与智能系统重点实验室(桂林理工大学), 广西桂林 541006

收稿日期:2022-05-23 修回日期:2022-09-01 接受日期:2022-09-05 发布日期:2022-09-23 出版日期:2023-06-10
通讯作者: 王真梅
作者简介:张奕（1977—），女，江西九江人，教授，博士，主要研究方向：生物信息学、机器学习、服务计算
王真梅（1996—），女，广西玉林人，硕士研究生，主要研究方向：生物信息学Email：1134573228@qq.com。
基金资助:
国家自然科学基金资助项目(62166014);广西自然科学基金资助项目(2020GXNSFAA297255);广西嵌入式技术与智能系统重点实验室项目(2019-01-16)

circRNA-disease association prediction by two-stage fusion on graph auto-encoder

Yi ZHANG¹^,², Zhenmei WANG¹()

^1.College of Information Science and Engineering，Guilin University of Technology，Guilin Guangxi 541006，China
^2.Guangxi Key Laboratory of Embedded Technology and Intelligent System （Guilin University of Technology），Guilin Guangxi 541006，China

Received:2022-05-23 Revised:2022-09-01 Accepted:2022-09-05 Online:2022-09-23 Published:2023-06-10
Contact: Zhenmei WANG
About author:ZHANG Yi， born in 1977， Ph. D.， professor. Her research interests include bioinformatics， machine learning， service computing.
Supported by:
National Natural Science Foundation of China(62166014);Guangxi Natural Science Foundation(2020GXNSFAA297255);Project of Guangxi Key Laboratory of Embedded Technology and Intelligent System(2019-01-16)

摘要/Abstract

摘要：

大部分现有的用于预测环状RNA（circRNA）与疾病之间关联关系的计算模型通常使用circRNA和疾病相关数据等生物学知识，配合已知的circRNA-疾病关联信息对来挖掘出潜在的关联信息。然而这些模型受已知关联构成的网络稀疏性、负样本过少等固有问题的影响，导致预测性能不佳。因此，在图自动编码器基础上引入归纳式矩阵补全及自注意力机制进行二阶段融合，以实现circRNA-疾病关联预测，由此构建的模型叫GIS-CDA （Graph auto-encoder combining Inductive matrix complementation and Self-attention mechanism for predicting CircRNA-Disease Association）。首先，计算circRNA集成和疾病集成的相似性，并利用图自动编码器学习circRNA和疾病的潜在特征，以获得低维表征；接着，将学习到的特征输入归纳式矩阵补全，以提高节点之间的相似性和依赖性；然后，将circRNA特征矩阵和疾病特征矩阵整合为circRNA-疾病特征矩阵，以增强预测的稳定性和精确性；最后，引入自注意力机制，从特征矩阵中提取重要特征，并减少对其他生物信息的依赖。五折交叉和十折交叉验证的结果显示：GIS-CDA获得的平均接收者操作特征曲线下面积（AUROC）值分别为0.930 3和0.939 3，前者比基于KATZ测度的人类circRNA-疾病关联预测模型（KATZHCDA）、基于深度矩阵分解方法的circRNA-疾病关联（DMFCDA）预测模型、RWR（Random Walk with Restart）和基于加速归纳式矩阵补全的circRNA-疾病关联（SIMCCDA）预测模型分别高出了13.19、35.73、13.28和5.01个百分点；GIS-CDA的精确率-召回率曲线下面积（AUPR）值分别为0.227 1和0.234 0，前者比上述对比模型分别高出了21.72、22.43、21.96和13.86个百分点。此外，在circRNADisease、circ2Disease和circR2Disease数据集上的消融实验和案例研究进一步验证了GIS-CDA在预测circRNA-疾病的潜在关联方面具有较好的性能。

关键词: 图自动编码器, 归纳式矩阵补全, 自注意力机制, 环状RNA, 环状RNA-疾病关联信息对

Abstract:

Most existing computational models for predicting associations between circular RNA （circRNA） and diseases usually use biological knowledge such as circRNA and disease-related data， and mine the potential association information by combining known circRNA-disease association information pairs. However， these models suffer from inherent problems such as sparsity and too few negative samples of networks composed of the known association， resulting in poor prediction performance. Therefore， inductive matrix completion and self-attention mechanism were introduced for two-stage fusion based on graph auto-encoder to achieve circRNA-disease association prediction， and the model based on the above is GIS-CDA （Graph auto-encoder combining Inductive matrix complementation and Self-attention mechanism for predicting CircRNA-Disease Association）. Firstly， the similarity of circRNA integration and disease integration was calculated， and graph auto-encoder was used to learn the potential features of circRNAs and diseases to obtain low-dimensional representations. Secondly， the learned features were input to inductive matrix complementation to improve the similarity and dependence between nodes. Thirdly， the circRNA feature matrix and disease feature matrix were integrated into circRNA-disease feature matrix to enhance the stability and accuracy of prediction. Finally， a self-attention mechanism was introduced to extract important features in the feature matrix and reduce the dependence on other biological information. The results of five-fold crossover and ten-fold crossover validation show that the Area Under Receiver Operating Characteristic curve （AUROC） values of GIS-CDA are 0.930 3 and 0.939 3 respectively， the former of which is 13.19，35.73，13.28 and 5.01 percentage points higher than those of the prediction models based on computational model of KATZ measures for Human CircRNA-Disease Association （KATZHCDA）， Deep Matrix Factorization for CircRNA-Disease Association （DMFCDA）， RWR （Random Walk with Restart） and Speedup Inductive Matrix Completion for CircRNA-Disease Associations （SIMCCDA）， respectively； the Area Under Precision-Recall curve （AUPR） values of GIS-CDA are 0.227 1 and 0.234 0 respectively， the former of which is 21.72， 22.43， 21.96 and 13.86 percentage points higher than those of the above comparison models respectively. In addition， ablation experiments and case studies on circRNADisease， circ2Disease and circR2Disease datasets， further validate the good performance of GIS-CDA in predicting the potential circRNA-disease association.

Key words: graph auto-encoder, inductive matrix completion, self-attention mechanism, circular RNA (circRNA), circRNA-disease association information pair

中图分类号:

TP183

张奕, 王真梅. 图自动编码器上二阶段融合实现的环状RNA-疾病关联预测[J]. 计算机应用, 2023, 43(6): 1979-1986.

Yi ZHANG, Zhenmei WANG. circRNA-disease association prediction by two-stage fusion on graph auto-encoder[J]. Journal of Computer Applications, 2023, 43(6): 1979-1986.

图/表 13

图1 GIS-CDA模型建立过程

Fig.1 Construction process of GISCDA model

图2 GIS-CDA交叉验证结果

Fig. 2 Cross validation results of GIS-CDA

表1 不同隐藏层维度d下的AUROC、AUPR值

Tab. 1 AUROC、AUPR values with different hidden layer dimension d

$d$	AUROC	AUPR
32	0.902 5	0.137 4
64	0.916 9	0.175 9
128	0.924 6	0.216 4
256	0.9303	0.2271
512	0.911 6	0.184 8

表1 不同隐藏层维度d下的AUROC、AUPR值

Tab. 1 AUROC、AUPR values with different hidden layer dimension d

$d$	AUROC	AUPR
32	0.902 5	0.137 4
64	0.916 9	0.175 9
128	0.924 6	0.216 4
256	0.9303	0.2271
512	0.911 6	0.184 8

表2 不同平衡系数α下的AUROC、AUPR值

Tab. 2 AUROC、AUPR values with different balance coefficient α

$α$	AUROC	AUPR
0.1	0.916 2	0.161 6
0.3	0.929 2	0.202 6
0.5	0.9303	0.2271
0.7	0.924 3	0.213 4
0.9	0.921 6	0.202 2

表2 不同平衡系数α下的AUROC、AUPR值

Tab. 2 AUROC、AUPR values with different balance coefficient α

$α$	AUROC	AUPR
0.1	0.916 2	0.161 6
0.3	0.929 2	0.202 6
0.5	0.9303	0.2271
0.7	0.924 3	0.213 4
0.9	0.921 6	0.202 2

表3 不同学习率l下的AUROC、AUPR值

Tab. 3 AUROC、AUPR values with different learning rate l

$l$	AUROC	AUPR
0.001	0.885 5	0.136 0
0.005	0.921 3	0.184 1
0.01	0.9303	0.2271
0.05	0.882 0	0.105 2
0.1	0.736 4	0.043 5

表3 不同学习率l下的AUROC、AUPR值

Tab. 3 AUROC、AUPR values with different learning rate l

$l$	AUROC	AUPR
0.001	0.885 5	0.136 0
0.005	0.921 3	0.184 1
0.01	0.9303	0.2271
0.05	0.882 0	0.105 2
0.1	0.736 4	0.043 5

表4 消融实验对比设置

Tab. 4 Comparison setting of ablation experiment

实验分组	自注意力机制	重建损失函数
第1组	引入	无
第2组	无	引入
第3组	无	无

图3 消融实验结果对比

Fig. 3 Comparison of ablation experimental results

表5 不同数据集数据细节

Tab. 5 Data details of different datasets

数据集名称	疾病数	circRNA数	关联数
circRNADisease	34	223	241
circ2Disease	46	215	240
circR2Disease	88	585	650

图4 不同数据集预测结果对比

Fig. 4 Comparison of prediction results on different datasets

图5 所提模型与现有模型的ROC曲线和PR曲线对比

Fig. 5 PR curve and ROC curve comparison of the proposed model and existing models

表6 所提模型与现有模型的AUROC值、AUPR值和运行时间对比

Tab.6 Comparison of AUROC values， AUPR values and running time of the proposed model and existing models

模型	时间/s	模型	时间/s
GIS-CDA	55	RWR	7 826
KATZHCDA	8 010	SIMCCDA	5
DMFCDA	116

表7 前10个与神经胶质癌相关的circRNA

Tab. 7 Top 10 circRNAs associated with glioma

排名	circRNA名称	PMID编号
1	circPVT1/hsa_circ_0001821	31865777
2	hsa_circRNA_100782/circHIPK3/ hsa_circ_0000284	32833501
3	hsa_circ_0001313/circCCDC66	34252882
4	circ-Foxo3/hsa_circ_0006404	33833780
5	Cir-ITCH/hsa_circ_0001141/ hsa_circ_001763	29887952
6	hsa_circRNA_103110/hsa_circ_103110/ hsa_circ_0004771	暂无证据
7	CDR1as/ciRS-7/hsa_circ_0001946	32894144
8	circPRKCI/hsa_circ_0067934	31409777
9	hsa_circRNA_102049	暂无证据
10	hsa_circ_0002495	暂无证据

表8 前10个与胃癌相关的circRNA

Tab. 8 Top 10 circRNAs associated with gastric cancer

排名	circRNA名称	PMID编号
1	CirITCH/hsa_circ_0001141/ hsa_circ_001763	33060778
2	hsa_circ_0001313/circCCDC66	32253030
3	hsa_circ_0007534	32419229
4	hsa_circRNA_103110/hsa_circ_103110/ hsa_circ_0004771	29098316
5	circPRKCI/hsa_circ_0067934	35113408
6	circ-Foxo3/hsa_circ_0006404	33833780
7	circZFR/hsa_circRNA_103809/hsa_circ_0072088	32572921
8	hsa_circRNA_102049	暂无证据
9	circSMARCA5/hsa_circ_0001445	30956729
10	circRNA_102913/hsa_circ_0058058	暂无证据

参考文献 32

1	SALZMAN J， CHEN R E， OLSEN M N， et al. Cell-type specific features of circular RNA expression［J］. PLoS Genetics， 2013， 9（12）： No.e1003777. 10.1371/journal.pgen.1003777
2	付瑶. 环状RNA——一个新的非编码RNA的功能与特性［J］. 吉林畜牧兽医， 2017， 38（11）：11-13. 10.3969/j.issn.1672-2078.2017.11.003
	FU Y. Function and properties of circular RNA — a new non-coding RNA［J］. Jilin Animal Husbandry and Veterinary Medicine， 2017， 38（11）：11-13. 10.3969/j.issn.1672-2078.2017.11.003
3	MENG S J， ZHOU H C， FENG Z Y， et al. circRNA： functions and properties of a novel potential biomarker for cancer ［J］. Molecular Cancer， 2017， 16： No.94. 10.1186/s12943-017-0663-2
4	张懿恋，张诗雨，胡吉，等. 环状RNA在糖尿病及其慢性并发症中的机制研究［J］. 中国医学科学院学报， 2022， 44（3）：521-528. 10.3881/j.issn.1000-503X.13436
	ZHANG Y L， ZHANG S Y， HU J， et al. Circular RNA in diabetes and its complications［J］. Acta Academiae Medicinae Sinicae， 2022， 44（3）：521-528. 10.3881/j.issn.1000-503X.13436
5	FAN C Y， LEI X J， WU F X. Prediction of circRNA-disease associations using KATZ model based on heterogeneous networks［J］. International Journal of Biological Sciences， 2018， 14（14）： 1950-1959. 10.7150/ijbs.28260
6	XIAO Q， YU H M， ZHONG J C， et al. An in-silico method with graph-based multi-label learning for large-scale prediction of circRNA-disease associations［J］. Genomics， 2020， 112（5）： 3407-3415. 10.1016/j.ygeno.2020.06.017
7	LEI X J， FANG Z Q， CHEN L N， et al. PWCDA： path weighted method for predicting circRNA-disease associations［J］. International Journal of Molecular Sciences， 2018， 19（11）： No.3410. 10.3390/ijms19113410
8	LEI X J， BIAN C. Integrating random walk with restart and k-nearest neighbor to identify novel circRNA-disease association ［J］. Scientific Reports， 2020， 10： No.1943. 10.1038/s41598-020-59040-0
9	YAN C， WANG J X， WU F X. DWNN-RLS： regularized least squares method for predicting circRNA-disease associations ［J］. BMC Bioinformatics， 2018， 19（S19）： No.520. 10.1186/s12859-018-2522-6
10	DING Y L， CHEN B L， LEI X J， et al. Predicting novel circRNA-disease associations based on random walk and logistic regression model［J］. Computational Biology and Chemistry， 2020， 87： No.107287. 10.1016/j.compbiolchem.2020.107287
11	WANG L， YOU Z H， HUANG Y A， et al. An efficient approach based on multi-sources information to predict circRNA-disease associations using deep convolutional neural network［J］. Bioinformatics， 2020， 36（13）： 4038-4046. 10.1093/bioinformatics/btz825
12	FAN C Y， LEI X J， PAN Y. Prioritizing circRNA-disease associations with convolutional neural network based on multiple similarity feature fusion［J］. Frontiers in Genetics， 2020， 11： No.540751. 10.3389/fgene.2020.540751
13	LI G H， WANG D C， ZHANG Y J， et al. Using graph attention network and graph convolutional network to explore human circrRNA-disease associations based on multi-source data［J］. Frontiers in Genetics， 2022， 13： No.829937. 10.3389/fgene.2022.829937
14	DEEPTHI K， JEREESH A S. Inferring potential circRNA-disease associations via deep autoencoder-based classification［J］. Molecular Diagnosis and Therapy， 2021， 25（1）： 87-97. 10.1007/s40291-020-00499-y
15	CHEN X， WANG L， QU J， et al. Predicting miRNA-disease association based on inductive matrix completion［J］. Bioinformatics， 2018， 34（24）： 4256-4265. 10.1093/bioinformatics/bty503
16	LU C Q， ZENG M， ZHANG F H， et al. Deep matrix factorization improves prediction of human circRNA-disease associations［J］. IEEE Journal of Biomedical and Health Informatics， 2021， 25（3）： 891-899. 10.1109/jbhi.2020.2999638
17	LI M L， LIU M Y， BIN Y N， et al. Prediction of circRNA-disease associations based on inductive matrix completion［J］. BMC Medical Genomics， 2020， 13（S5）： No.42. 10.1186/s12920-020-0679-0
18	WANG D， WANG J， LU M， et al. Inferring the human microRNA functional similarity and functional network based on microRNA-associated diseases ［J］. Bioinformatics， 2010， 26（13）： 1644-1650. 10.1093/bioinformatics/btq241
19	LI Z W， LI J S， NIE R， et al. A graph auto-encoder model for miRNA-disease associations prediction［J］. Briefings in Bioinformatics， 2021， 22（4）： No.bbaa240. 10.1093/bib/bbaa240
20	LI J， ZHANG S， LIU T， et al. Neural inductive matrix completion with graph convolutional networks for miRNA-disease association prediction ［J］. Bioinformatics， 2020， 36（8）： 2538-2546. 10.1093/bioinformatics/btz965
21	BIAN C， LEI X J， WU F X. GATCDA： predicting circRNA-disease associations based on graph attention network［J］. Cancers， 2021， 13（11）： No.2595. 10.3390/cancers13112595
22	JIN C， SHI Z W， LIN K， et al. Predicting miRNA-disease association based on neural inductive matrix completion with graph autoencoders and self-attention mechanism［J］. Biomolecules， 2022， 12（1）： No.64. 10.3390/biom12010064
23	FAN C Y， LEI X J， FANG Z Q， et al. circR2Disease： a manually curated database for experimentally supported circular RNAs associated with various diseases ［J］. Database， 2018， 2018： No.bay044. 10.1093/database/bay044
24	ZHAO Z， WANG K Y， WU F， et al. circRNA disease： a manually curated database of experimentally supported circRNA-disease associations［J］. Cell Death and Disease， 2018， 9： No.475. 10.1038/s41419-018-0503-3
25	YAO D X， ZHANG L， ZHENG M Y， et al. circ2Disease： a manually curated database of experimentally validated circRNAs in human disease ［J］. Scientific Reports， 2018， 8： No.11018. 10.1038/s41598-018-29360-3
26	VURAL H， KAYA M， ALHAJJ R. A model based on random walk with restart to predict circRNA-disease associations on heterogeneous network ［C］// Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining. New York： ACM， 2019： 929-932. 10.1145/3341161.3343514
27	ZHU J L， YE J L， ZHANG L， et al. Differential expression of circular RNAs in glioblastoma multiforme and its correlation with prognosis ［J］. Translational Oncology， 2017， 10（2）： 271-279. 10.1016/j.tranon.2016.12.006
28	DANG Y， LAN F H， OUYANG X J， et al. Expression and clinical significance of long non-coding RNA HNF1A-AS1 in human gastric cancer ［J］. World Journal of Surgical Oncology， 2015， 13： No.302. 10.1186/s12957-015-0706-3
29	CHI G N， YANG F W， XU D H， et al. Silencing hsa_circ_PVT1 （circPVT1） suppresses the growth and metastasis of glioblastoma multiforme cells by up-regulation of miR-199a-5p ［J］. Artificial Cells， Nanomedicine， and Biotechnology， 2020， 48（1）：188-196. 10.1080/21691401.2019.1699825
30	YIN H Q CUI X. Knockdown of circHIPK3 facilitates temozolomide sensitivity in glioma by regulating cellular behaviors through miR-524-5p/KIF2A-mediated PI3K/AKT pathway［J］. Cancer Biotherapy and Radiopharmaceuticals， 2021， 36（7）： 556-567. 10.1089/cbr.2020.3575
31	PENG Y， WANG H H. Cir-ITCH inhibits gastric cancer migration， invasion and proliferation by regulating the Wnt/β-catenin pathway ［J］. Scientific Reports， 2020， 10： No.17443. 10.1038/s41598-020-74452-8
32	ZHANG Q， MIAO Y C， FU Q S， et al. CircRNACCDC66 regulates cisplatin resistance in gastric cancer via the miR-618/BCL2 axis［J］. Biochemical and Biophysical Research Communications， 2020， 526（3）： 713-720. 10.1016/j.bbrc.2020.03.156

[1]	孙浩, 曹健, 李海生, 毛典辉. 基于改进胶囊网络的会话型推荐模型[J]. 《计算机应用》唯一官方网站, 2023, 43(4): 1043-1049.
[2]	左亚尧, 陈皓宇, 陈致然, 洪嘉伟, 陈坤. 融合多语义特征的命名实体识别方法[J]. 《计算机应用》唯一官方网站, 2022, 42(7): 2001-2008.
[3]	任炜, 白鹤翔. 基于全局与局部标签关系的多标签图像分类方法[J]. 《计算机应用》唯一官方网站, 2022, 42(5): 1383-1390.
[4]	刘志华, 陈文洁, 陈爱斌. 基于自注意力机制时频谱同源特征融合的鸟鸣声分类[J]. 《计算机应用》唯一官方网站, 2022, 42(4): 1260-1268.
[5]	赵玉明, 顾慎凯. 融合残差密集块自注意力机制和生成对抗网络的对抗攻击防御模型[J]. 《计算机应用》唯一官方网站, 2022, 42(3): 921-929.
[6]	李亚鸣, 邢凯, 邓洪武, 王志勇, 胡璇. 基于小样本无梯度学习的卷积结构预训练模型性能优化方法[J]. 《计算机应用》唯一官方网站, 2022, 42(2): 365-374.
[7]	耿艳兵, 廉永健. 基于多粒度特征生成对抗网络的跨分辨率行人重识别[J]. 《计算机应用》唯一官方网站, 2022, 42(11): 3573-3579.
[8]	席荣康, 蔡满春, 芦天亮, 李彦霖. 基于自注意力机制和时空特征的Tor网站流量分析模型[J]. 《计算机应用》唯一官方网站, 2022, 42(10): 3084-3090.
[9]	党伟超, 李涛, 白尚旺, 高改梅, 刘春霞. 基于自注意力长短期记忆网络的Web软件系统实时剩余寿命预测方法[J]. 计算机应用, 2021, 41(8): 2346-2351.
[10]	陈佳伟, 韩芳, 王直杰. 基于自注意力门控图卷积网络的特定目标情感分析[J]. 计算机应用, 2020, 40(8): 2202-2206.
[11]	李生武, 张选德. 基于自注意力机制的多域卷积神经网络的视觉追踪[J]. 计算机应用, 2020, 40(8): 2219-2224.
[12]	张小川, 戴旭尧, 刘璐, 冯天硕. 融合多头自注意力机制的中文短文本分类模型[J]. 计算机应用, 2020, 40(12): 3485-3489.
[13]	许一宁, 何小海, 张津, 卿粼波. 基于多层次分辨率递进生成对抗网络的文本生成图像方法[J]. 计算机应用, 2020, 40(12): 3612-3617.

图自动编码器上二阶段融合实现的环状RNA-疾病关联预测

circRNA-disease association prediction by two-stage fusion on graph auto-encoder

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 32

相关文章 13

编辑推荐

Metrics