基于卷积神经网络多尺度特征的大豆基因组表型预测

doi:10.7523/j.ucas.2023.046

摘要/Abstract

摘要： 在育种中，常常通过利用单核苷酸多态性(SNPs)来预测表型以辅助育种，提高育种效率。传统的统计分析方法受到数据缺失等诸多因素的限制，在一些情况下效果不佳。针对此问题，提出一种利用多尺度特征进行植物性状预测的卷积神经网络模型(MSF-CNN)，该模型通过卷积提取3个不同尺度的SNPs特征，对植物性状数值进行回归预测，并通过对模型中SNPs的权重分析SNP位点的显著性。测试结果表明，与目前已知的其他方法相比，MSF-CNN模型在有基因型数据缺失值的数据集上表型预测的准确性更高。此外，通过显著性图研究基因型对性状的贡献，发现数个较显著的SNP位点。说明该深度学习模型可以更准确地预测定量表型，并能够高效识别与全基因组关联研究相关的SNP位点。

关键词: 遗传筛选, 深度学习, 全基因组关联分析, 大豆

Abstract: In breeding, single nucleotide polymorphisms (SNPs) in the genome are often used to predict quantitative phenotypes to assist breeding, thereby improving breeding efficiency. The traditional statistical analysis method is limited by many factors including missing data, and its performance sometimes can not meet the requirements. In this paper, we proposed a multi-scale feature convolutional neural network model (MSF-CNN) to predict plant traits. The model extracted SNP features at three different scales through convolution and analyzed the significance of SNP sites through the weight of the SNPs input into the model. The test results showed that MSF-CNN model performed with higher accuracy than the known methods and other deep learning models in phenotype prediction on the datasets with missing genotypic data. This paper also studied the contribution of genotype to traits through saliency map, and discovered several significant SNP loci. These results showed that, compared with other known methods available at present, the deep learning model proposed in this paper can obtain more accurate prediction results of quantitative phenotypes, and can also effectively and efficiently identify SNPs associated with genome-wide association research.

Key words: gene selection, deep learning, genome-wide association study, soybean

中图分类号:

Q943.2

林昱彤, 王红, 柴团耀. 基于卷积神经网络多尺度特征的大豆基因组表型预测[J]. 中国科学院大学学报, 2024, 41(4): 468-476.

LIN Yutong, WANG Hong, CHAI Tuanyao. Multi-scale featured convolution neural network-based soybean phenotypic prediction[J]. Journal of University of Chinese Academy of Sciences, 2024, 41(4): 468-476.

参考文献

[1] Zhao Y S, Gowda M, Liu W X, et al. Accuracy of genomic selection in European maize elite breeding populations[J]. Theoretical and Applied Genetics, 2012, 124(4): 769-776. DOI: 10.1007/s00122-011-1745-y.
[2] Spindel J, Begum H, Akdemir D, et al. Genomic selection and association mapping in rice (Oryza sativa): effect of trait genetic architecture, training population composition, marker number and statistical model on accuracy of rice genomic selection in elite, tropical rice breeding lines[J]. PLoS Genetics, 2015, 11(2): e1004982. DOI: 10.1371/journal.pgen.1004982.
[3] Xavier A, Jarquin D, Howard R, et al. Genome-wide analysis of grain yield stability and environmental interactions in a multiparental soybean population[J]. G3-Genes Genomes Genetics, 2018, 8(2): 519-529. DOI: 10.1534/g3.117.300300.
[4] Endelman J B. Ridge regression and other kernels for genomic selection with R package rrBLUP[J]. The Plant Genome, 2011, 4(3): 250-255. DOI: 10.3835/plantgenome2011.08.0024.
[5] Wang J X, Joshi T, Valliyodan B, et al. A Bayesian model for detection of high-order interactions among genetic variants in genome-wide association studies[J]. BMC Genomics, 2015, 16: 1011. DOI: 10.1186/s12864-015-2217-6.
[6] Rutkoski J E, Poland J, Jannink J L, et al. Imputation of unordered markers and the impact on genomic selection accuracy[J]. G3 Genes | Genomes | Genetics, 2013, 3(3): 427-439. DOI: 10.1534/g3.112.005363.
[7] Sapoval N, Aghazadeh A, Nute M G, et al. Current progress and open challenges for applying deep learning across the biosciences[J]. Nature Communications, 2022, 13(1): 1-12. DOI: 10.1038/s41467-022-29268-7.
[8] Tong L, Mitchel J, Chatlin K, et al. Deep learning based feature-level integration of multi-omics data for breast cancer patients survival analysis[J]. BMC Medical Informatics and Decision Making, 2020, 20(1): 225. DOI: 10.1186/s12911-020-01225-8.
[9] Uppu S, Krishna A, Gopalan R P. A deep learning approach to detect SNP interactions[J]. Journal of Software, 2016, 11(10): 965-975. DOI: 10.17706/jsw.11.10.965-975.
[10] Liang Z H, Huang J X, Zeng X, et al. DL-ADR: a novel deep learning model for classifying genomic variants into adverse drug reactions[J]. BMC Medical Genomics, 2016, 9(S2): 48. DOI: 10.1186/s12920-016-0207-4.
[11] Lee G, Nho K, Kang B, et al. Predicting Alzheimer’s disease progression using multi-modal deep learning approach[J]. Scientific Reports, 2019, 9(1): 1-12. DOI: 10.1038/s41598-018-37769-z.
[12] Zingaretti L M, Gezan S A, Ferrão L F V, et al.Exploring deep learning for complex trait genomic prediction in polyploid outcrossing species[J]. Frontiers in Plant Science, 2020, 11: 25. DOI: 10.3389/fpls.2020.00025.
[13] Whalen S, Schreiber J, Noble W S, et al. Navigating the pitfalls of applying machine learning in genomics[J]. Nature Reviews Genetics, 2022, 23(3): 169-181. DOI: 10.1038/s41576-021-00434-9.
[14] Xavier A, Beavis W D, Specht J E, et al. SoyNAM: soybean nested association mapping dataset[DB]. R package version, 2015, 1.
[15] Song Q J, Yan L, Quigley C, et al. Genetic characterization of the soybean nested association mapping population[J]. The Plant Genome, 2017, 10(2): 10.3835/plantgenome,2016. 10.0109. DOI: 10.3835/plantgenome2016.10.0109.
[16] Li Y, Willer C J, Ding J, et al. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes[J]. Genetic Epidemiology, 2010, 34(8): 816-834. DOI: 10.1002/gepi.20533.
[17] Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). July 21-26, 2017, Honolulu, HI, USA. IEEE, 2017: 2261-2269. DOI: 10.1109/CVPR.2017.243.
[18] Zhang X Z, Wang Y F, Shi W S. pCAMP: performance comparison of machine learning packages on the edges[EB/OL]. (2019-06-05) [2023-03-24]. https://arxiv.org/abs/1906.01878.
[19] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[EB/OL]. (2014-09-04) [2023-03-24]. https://arxiv.org/abs/1409.1556.
[20] Srivastava N, Hinton G, Krizhevsky A, et al. Dropout: a simple way to prevent neural networks from overfitting[J]. The Journal of Machine Learning Research, 2014, 15(1), 1929-1958.
[21] Liu Y, Wang D L, He F, et al. Phenotype prediction and genome-wide association study using deep convolutional neural network of soybean[J]. Frontiers in Genetics, 2019, 10: 1091. DOI: 10.3389/fgene.2019.01091.
[22] Schmutz J, Cannon S B, Schlueter J, et al. Erratum: genome sequence of the palaeopolyploid soybean[J]. Nature, 2010, 465(7294): 120. DOI: 10.1038/nature08957.
[23] Grant D, Nelson R T, Cannon S B, et al. SoyBase, the USDA-ARS soybean genetics and genomics database[J]. Nucleic Acids Research, 2010, 38(Suppl 1): D843-D846. DOI: 10.1093/nar/gkp798.
[24] Joshi T, Patil K, Fitzpatrick M R, et al. Soybean knowledge base (SoyKB): a web resource for soybean translational genomics[J]. BMC Genomics, 2012, 13(Suppl 1): S15. DOI: 10.1186/1471-2164-13-S1-S15.
[25] Bateman A, Coin L, Durbin R, et al. The Pfam protein families database[J]. Nucleic Acids Research, 2004, 32 (Suppl 1): D138-D141. DOI: 10.1093/nar/gkh121.
[26] Xu Z Y, Wang R K, Kong K K, et al. An APETALA2/ethylene responsive factor transcription factor GmCRF4a regulates plant height and auxin biosynthesis in soybean[J]. Frontiers in Plant Science, 2022, 13: 983650. DOI: 10.3389/fpls.2022.983650.