基于全血基因表达谱的脑组织基因表达量预测模型的建立

doi:10.3969/j.issn.1006-7795.2019.05.013

摘要/Abstract

摘要： 目的基因表达分析是阐释生物学表型和辅助疾病诊断的有力工具，脑组织的基因表达实验样品采集工作难度大、风险高、成本高，亟需一种能规避脑组织取样的表达谱检测替代方案。方法以基因型－组织表达数据库（genotype-tissue expression，GTEx）中的脑组织样本配对的全血基因表达谱为输入特征数据，以13个脑组织的基因表达量分别为目标数据，挖掘全血基因表达量与脑组织中任一基因表达量数值的多对多关联关系，进而构建一个基于全血基因表达谱的未取样脑组织中基因表达量的回归预测模型。结果针对每个基因分别提取包含15个最相关的全血基因表达量特征构成低维度的新特征数据集，构建了13个脑组织所有基因表达量线性回归预测模型。预测模型平均绝对误差为0.406~0.542，均方根误差为0.558~0.941。结论本研究提出了一种基于全血基因表达量数据的脑组织基因表达量预测模型，证明仅用全血表达谱数据能比较准确地预测出未取样脑组织基因表达量，有望在转录组研究中规避脑组织样本的手术取样，为脑组织相关疾病的基因表达谱研究提供一种备选工具。

关键词: 脑组织, 全血, 基因表达, 预测模型, 特征选择

Abstract: Objective Gene expression analysis is a powerful tool for explaining biological phenotypes and assisting disease diagnosis. It is difficult, risky and expensive to collect the samples of gene expression experiment of brain tissue. It is urgent to find an alternative method to detect the expression profiles of brain tissue from other available samples. Methods Using the whole blood gene expression profiles matched with brain tissue samples from GTEx (Genotype-Tissue Expression) database as input features and 13 brain tissue expression profiles as targets, we mined many-to-many correlations between the gene expression level in whole blood and that of specific brain tissue. Then we construct a predictive regression model of gene expression level of unavailable brain tissue based on the expression level of whole blood gene expression profiles. Results A new low-dimensional feature dataset for each gene in each brain tissue was constructed by extracting 15 most relevant gene expression features from whole blood, and a linear regression prediction model for all gene in 13 brain tissues was constructed. The mean absolute error (MAE) of the prediction model is between 0.406 and 0.542, and the root mean square error (RMSE) is between 0.558 and 0.941. Conclusion A prediction model of gene expression in brain tissues based on whole blood gene expression profile is proposed. It is proved that the gene expression in unsampled brain tissue can be predicted relatively accurately only by using whole blood expression profile data. It is possible that the surgical sampling of brain tissue samples can be avoided in transcriptome research, thus providing an alternative for the study of gene expression profiles of brain tissue-related diseases.

Key words: brain tissue, whole blood, gene expression, prediction model, feature selection

中图分类号:

徐文剑, 李巍. 基于全血基因表达谱的脑组织基因表达量预测模型的建立[J]. 首都医科大学学报, 2019, 40(5): 731-737.

Xu Wenjian, Li Wei. Building a prediction model of brain tissues gene expression based on whole blood gene expression profiles[J]. Journal of Capital Medical University, 2019, 40(5): 731-737.

参考文献

[1] Van't Veer L J, Dai H, Van De Vijver M J, et al. Gene expression profiling predicts clinical outcome of breast cancer[J]. Nature, 2002, 415(6871):530-536.
[2] Bullinger L, Dohner K, Bair E, et al. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia[J]. N Engl J Med, 2004, 350(16):1605-1616.
[3] Finak G, Bertos N, Pepin F, et al. Stromal gene expression predicts clinical outcome in breast cancer[J]. Nat Med, 2008, 14(5):518-527.
[4] Byron S A, Van Keuren-Jensen K R, Engelthaler D M, et al. Translating RNA sequencing into clinical diagnostics:opportunities and challenges[J]. Nat Rev Genet, 2016, 17(5):257-271.
[5] Costa V, Aprile M, Esposito R, et al. RNA-Seq and human complex diseases:recent accomplishments and future perspectives[J]. Eur J Hum Genet, 2013, 21(2):134-142.
[6] Berent D, Macander M, Szemraj J, et al. Vascular endothelial growth factor a gene expression level is higher in patients with major depressive disorder and not affected by cigarette smoking, hyperlipidemia or treatment with statins[J]. Acta Neurobiol Exp (Wars), 2014, 74(1):82-90.
[7] GTEx Consortium. The genotype-tissue expression (GTEx) project[J]. Nat Genet, 2013, 45(6):580-585.
[8] Koleti A, Terryn R, Stathias V, et al. Data portal for the library of integrated network-based cellular signatures (LINCS) program:integrated access to diverse large-scale cellular perturbation response data[J]. Nucleic Acids Res, 2018, 46(D1):D558-D566.
[9] Wang J, Gamazon E R, Pierce B L, et al. Imputing gene expression in uncollected tissues within and beyond GTEx[J]. Am J Hum Genet, 2016, 98(4):697-708.
[10] Voineagu I. Gene expression studies in autism:moving from the genome to the transcriptome and beyond[J]. Neurobiol Dis, 2012, 45(1):69-75.
[11] Ch'ng C, Kwok W, Rogic S, et al. Meta-analysis of gene expression in autism spectrum disorder[J]. Autism Res, 2015, 8(5):593-608.
[12] Gregg J P, Lit L, Baron C A, et al. Gene expression changes in children with autism[J]. Genomics, 2008, 91(1):22-29.
[13] Quesnel-Vallieres M, Weatheritt R J, Cordes S P, et al. Autism spectrum disorder:insights into convergent mechanisms from transcriptomics[J]. Nat Rev Genet, 2019, 20(1):51-63.
[14] Mundalil Vasu M, Anitha A, Thanseem I, et al. Serum microRNA profiles in children with autism[J]. Mol Autism, 2014, 5:40.
[15] Sullivan P F, Fan C, Perou C M. Evaluating the comparability of gene expression in blood and brain[J]. Am J Med Genet B Neuropsychiatr Genet, 2006, 141B(3):261-268.
[16] Bosker F J, Gladkevich A V, Pietersen C Y, et al. Comparison of brain and blood gene expression in an animal model of negative symptoms in schizophrenia[J]. Prog Neuropsychopharmacol Biol Psychiatry, 2012, 38(2):142-148.
[17] Hensman Moss D J, Flower M D, Lo K K, et al. Huntington's disease blood and brain show a common gene expression pattern and share an immune signature with Alzheimer's disease[J]. Sci Rep, 2017, 7:44849.
[18] Halloran J W, Zhu D, Qian D C, et al. Prediction of the gene expression in normal lung tissue by the gene expression in blood[J]. BMC Med Genomics, 2015, 8:77.
[19] Teipel S J, Grothe M J, Metzger C D, et al. Robust detection of impaired resting state functional connectivity networks in alzheimer's disease using elastic net regularized regression[J]. Front Aging Neurosci, 2017, 8:318.
[20] Rahmanl R, Perera C, Ghosh S, et al. Adaptive multi-task elastic net based feature selection from pharmacogenomics databases[J]. Conf Proc IEEE Eng Med Biol Soc, 2018, 2018:279-282.
[21] Marafino B J, Boscardin W J, Dudley R A. Efficient and sparse feature selection for biomedical text classification via the elastic net:application to ICU risk stratification from nursing notes[J]. J Biomed Inform, 2015, 54:114-120.
[22] Glatt S J, Everall I P, Kremen W S, et al. Comparative gene expression analysis of blood and brain provides concurrent validation of SELENBP1 up-regulation in schizophrenia[J]. Proc Natl Acad Sci U S A, 2005, 102(43):15533-15538.
[23] Chen H, Wang N, Zhao X, et al. Gene expression alterations in bipolar disorder postmortem brains[J]. Bipolar Disord, 2013, 15(2):177-187.
[24] Mahajan G J, Vallender E J, Garrett M R, et al. Altered neuro-inflammatory gene expression in hippocampus in major depressive disorder[J]. Prog Neuropsychopharmacol Biol Psychiatry, 2018, 82:177-186.
[25] Jansen R, Penninx B W, Madar V, et al. Gene expression in major depressive disorder[J]. Mol Psychiatry, 2016, 21(3):444.
[26] Warden A S, Mayfield R D. Gene expression profiling in the human alcoholic brain[J]. Neuropharmacology, 2017, 122(161-174.
[27] Lewis P A, Cookson M R. Gene expression in the Parkinson's disease brain[J]. Brain Res Bull, 2012, 88(4):302-312.
[28] Liu B, Wei Y, Zhang Y, et al. Deep neural networks for high dimension, low sample size data[C]. Proceedings of the twenty-sixth international joint conference on artificial intelligence (IJCAI-17),Melbourne, 2017.