基于变分自编码器神经网络的脑卒中药物重定位研究

Drugs Repositioning for Ischemic Stroke Based on Variational Autoencoder Neural Network

摘要

目的 构建基于变分自编码器的神经网络模型，预测不同药物作用机制的相似性，支持高效筛选脑卒中候选治疗药物，提高药物研发效率。方法 设计双编码器-解码器对比学习神经网络模型，整合药物作用下的单细胞转录组测序（RNA-seq）数据，构建药物扰动基因表达数据训练集用于模型训练；将药物表型效应特征映射至双曲潜在空间；基于潜在空间中药物特征表示的距离进行药物重定位，筛选出候选药物；通过生物信息学分析对筛选结果进行验证。结果 构建的双曲对比学习变分自编码器（hyperbolic contrastive variational autoencoder，HCVAE）模型，在聚类效果上显著优于标准变分自编码器（variational autoencoder，VAE）模型，能有效提取药物表型特征；通过模型推断筛选出赛度替尼等与依达拉奉具有相似效应的候选药物。进一步分析表明，赛度替尼的表型效应与其对细胞凋亡过程中受体相互作用蛋白激酶1、caspase等关键蛋白的调控相关。结论 HCVAE模型能有效提取“药物-疾病”表型特征，筛选出与依达拉奉效应具有高度相似性的缺血性脑卒中候选治疗药物赛度替尼，该模型可作为脑卒中药物筛选的高效平台，同时为药物重定位研究提供新思路。

Abstract

Objective This study aimed to predict the similarity in the mechanisms of action for different drugs, facilitate the efficient screening of candidate therapeutics for ischemic stroke, and enhance drug development efficiency through a modified variational autoencoder-based neural network. Methods A dual encoder-decoder variational autoencoder (VAE), termed hyperbolic contrastive VAE (HCVAE), was developed to extract features from training datasets comprising integrated drug-perturbed single-cell RNA sequencing (RNA-seq) gene expression profiles. Phenotypic effect features of drugs were mapped into a hyperbolic latent space. Drug repositioning was performed based on the distances between drug feature representations in the latent space to identify candidate compounds. Biological effects were preliminarily validated through bioinformatics analyses.Results The constructed HCVAE model demonstrated superior performance compared to standard VAEs in effectively extracting drug phenotypic features. Cerdulatinib was identified as a candidate therapeutic exhibiting effects comparable to edaravone in ischemic stroke models. Further analysis indicated that cerdulatinib,s phenotypic effects are associated with the regulation of key proteins RIPK1 and caspases within the cellular apoptosis pathway. Conclusion The HCVAE model effectively extracts “drug-disease” phenotypic features and can serve as a robust platform for candidate drug screening, providing novel insights for drug repositioning research.

关键词

变分自编码器 / 对比学习 / 赛度替尼 / 依达拉奉 / 脑卒中

Key words

variational autoencoder / contrastive learning / cerdulatinib / edaravone / ischemic stroke

参考文献

[1] Eltzschig H K, Eckle T. Ischemia and reperfusion: from mechanism to translation[J].Nat Med, 2011, 17(11): 1391-1401.

[2] Qin C, Yang S, Chu Y H,et al. Signaling pathways involved in ischemic stroke: molecular mechanisms and therapeutic interventions[J].Signal Transduct Target Ther, 2022, 7(1): 215.

[3] Tian J, Guo S, Chen H,et al. Combination of emricasan with ponatinib synergistically reduces ischemia/reperfusion injury in rat brain through simultaneous prevention of apoptosis and necroptosis[J].Transl Stroke Res, 2018, 9(4): 382-392.

[4] Peng Z M, Zhang Y Y, Wei D,et al. MALT1 promotes necroptosis in stroke rat brain viatargeting the A20/RIPK3 pathway[J].Arch Biochem Biophys, 2023, 735: 109502.

[5] Zhang Y Y, Liu W N, Li Y Q,et al. Ligustroflavone reduces necroptosis in rat brain after ischemic stroke through targeting RIPK1/RIPK3/MLKL pathway[J].Naunyn Schmiedebergs Arch Pharmacol, 2019, 392(9): 1085-1095.

[6] Tian J, Zhang Y Y, Peng Y W,et al. Polymyxin B reduces brain injury in ischemic stroke rat through a mechanism involving targeting ESCRT—III machinery and RIPK1/RIPK3/MLKL pathway[J].J Cardiovasc Transl Res, 2022, 15(5): 1129-1142.

[7] Davaanyam D, Lee H, Seol S I,et al. HMGB1 induces hepcidin upregulation in astrocytes and causes an acute iron surge and subsequent ferroptosis in the postischemic brain[J].Exp Mol Med, 2023, 55(11): 2402-2416.

[8] Peng J, He J, Lin L,et al. Neural stem cell extracellular vesicles carrying YBX1 inhibited neuronal pyroptosis through increasing m6A—modified GPR30 stability and expression in ischemic stroke[J].Transl Stroke Res, 2025, 16(2): 262-279.

[9] Duan Q, Reid S P, Clark N R,et al. L1000CDS2: LINCS L1000 characteristic direction signatures search engine[J].NPJ Syst Biol Appl, 2016, 2: 16015.

[10] Butler A, Hoffman P, Smibert P,et al. Integrating single—cell transcriptomic data across different conditions, technologies, and species[J].Nat Biotechnol, 2018, 36(5): 411-420.

[11] Kiselev V Y, Andrews T S, Hemberg M. Challenges in unsupervised clustering of single—cell RNA—seq data[J].Nat Rev Genet, 2019, 20(5): 273-282.

[12] Svensson V, Vento—Tormo R, Teichmann S A. Exponential scaling of single—cell RNA—seq in the past decade[J].Nat Protoc, 2018, 13(4): 599-604.

[13] Eraslan G, Simon L M, Mircea M,et al. Single—cell RNA—seq denoising using a deep count autoencoder[J].Nat Commun, 2019, 10: 390.

[14] Lopez R, Regier J, Cole M B,et al. Deep generative modeling for single—cell transcriptomics[J].Nat Methods, 2018, 15(12): 1053-1058.

[15] Kwon O S, Lee H, Kong H J,et al. Connectivity map—based drug repositioning of bortezomib to reverse the metastatic effect of GALNT14 in lung cancer[J].Oncogene, 2020, 39(23): 4567-4580.

[16] Keenan A B, Wojciechowicz M L, Wang Z,et al. Connectivity mapping: methods and applications[J].Annu Rev Biomed Data Sci, 2019, 2: 69-92.

[17] Zhao Y, Chen X, Chen J,et al. Decoding connectivity map—based drug repurposing for oncotherapy[J].Brief Bioinform, 2023, 24(3): bbad142.

[18] Subramanian A, Narayan R, Corsello S M,et al. A next generation connectivity map: L1000 platform and the first 1,000,000 profiles[J].Cell, 2017, 171(6): 1437-1452.e17.

[19] Lin K, Li L, Dai Y,et al. A comprehensive evaluation of connectivity methods for L1000 data[J].Brief Bioinform, 2020, 21(6): 2194-2205.

[20] Peng J J, Zhang Y Y, Li R F,et al. Hybrid approach for drug—target interaction predictions in ischemic stroke models[J].Artif Intell Med, 2025, 161: 103067.

[21] Kingma D P, Welling M. Auto—encoding variational bayes[EB/OL]. (2022—11—10) [2025—09—07].https://arxiv.org/abs/1312.6114.

[22] Bengio Y, Courville A, Vincent P. Representation learning: a review and new perspectives[J].IEEE Trans Pattern Anal Mach Intell, 2013, 35(8): 1798-1828.

[23] Weinberger E, Lin C, Lee S I. Isolating salient variations of interest in single—cell transcriptomic data with contrastiveVI[J].Nat Methods, 2023, 20(9): 1336-1345.

[24] Yang M, Yang Y, Xie C,et al. Contrastive learning enables rapid mapping to multimodal single—cell atlas of multimillion scale[J].Nat Mach Intell, 2022, 4(8): 696-709.

[25] Hinton G, Vinyals O, Dean J. Distilling the knowledge in a neural network[EB/OL]. (2015—05—09) [2025—09—07].https://arxiv.org/abs/1503.02531.

[26] van den Oord A, Li Y, Vinyals O. Representation learning with contrastive predictive coding[EB/OL]. (2019—01—22)[2025—09—07].https://arxiv.org/abs/1807.03748.

[27] Duan Q, Reid S, Clark N,et al. L1000CDS2: LINCS L1000 characteristic direction signatures search engine[EB/OL]. (2025—03—31)[2025—09—07].http://amp.pharm.mssm.edu/static/hdfs/l1000cds2/cpcd-gse70138.bson.

[28] Schnekenburger M, Dicato M, Diederich M F. Anticancer potential of naturally occurring immunoepigenetic modulators: a promising avenue?[J].Cancer, 2019, 125(10): 1612-1628.

[29] Wang C, Huang M, Lin Y,et al. ENO₂—derived phosphoenolpyruvate functions as an endogenous inhibitor of HDAC1 and confers resistance to antiangiogenic therapy[J].Nat Metab, 2023, 5(10): 1765-1786.

[30] Cozzolino F, Iacobucci I, Monaco V,et al. Lysines acetylome and methylome profiling of H3 and H4 histones in trichostatin A: treated stem cells[J].Int J Mol Sci, 2021, 22(4): 2063.

[31] Lu Z P, Ju Z L, Shi G Y,et al. Histone deacetylase inhibitor trichostatin A reduces anti—DNA autoantibody production and represses IgH gene transcription[J].Biochem Biophys Res Commun, 2005, 330(1): 204-209.

[32] Ma J, Xing W, Coffey G,et al. Cerdulatinib, a novel dual SYK/JAK kinase inhibitor, has broad anti—tumor activity in both ABC and GCB types of diffuse large B cell lymphoma[J].Oncotarget, 2015, 6(41): 43881-43896.

[33] Porter A G, Jänicke R U. Emerging roles of caspase—3 in apoptosis[J].Cell Death Differ, 1999, 6(2): 99-104.

[34] Ai X, Smith M C, Feltus F A. Generative adversarial networks applied to gene expression analysis: an interdisciplinary perspective[J].Comput Syst Oncol, 2023, 3(3): e1050.

[35] Kim J, Seok J. ctGAN: combined transformation of gene expression and survival data with generative adversarial network[J]. Brief Bioinform, 2024, 25(4): bbae325.

[36] Lee M. Recent advances in generative adversarial networks for gene expression data: a comprehensive review[J].Mathematics, 2023, 11(14): 3055.

基金资助

国家自然科学基金资助项目(82173815)

国家自然科学基金资助项目(81872873)