Abstract：With the high prevalence of neurological diseases, neural stem cell therapy has emerged as a significant research focus. However, current technologies still face challenges in accurately evaluating the neural stem cell differentiation process. This work aimed to develop a method using convolutional neural networks (CNNs) with transfer learning to efficiently classify images of label-free neuronal cells at different differentiation stages. Additionally, single-cell transcriptome data was leveraged to enhance the accuracy of cell differentiation stage detection and to validate the method′s effectiveness. The dataset including1 026 bright-field images of PC12 cells across different differentiation stages, supplemented by 100 images of HEK-293A or HT1080 cells obtained from publicly available online databases. Four CNNs including ResNet34, were evaluated. The models were conducted pre-training on the ImageNet dataset, followed by fine-tuning using a cell image dataset with annotation. The transcriptomic analysis was conducted on each cell after classifying the differentiation stages of thebright-field neural cell images. The results showed that the ResNet34-TL model performed best in classifying neural cells with an accuracy of 95.8%. The transcriptomic analysis of cells classified as undifferentiated, low-differentiated, and highly-differentiated by ResNet34-TL model indicates significant differences, particularly between the undifferentiated and highly-differentiated groups. Cells in the low-differentiated group exhibited characteristics of both extremes, suggesting a transitional state. The differential expression level of characteristic genes across the three groups revealed that the transcriptome-based classification was consistent with the model's classification results. ResNet34-TL exhibited strong generalization ability in the analysis of neuronal cell differentiation stages, and its capacity to effectively distinguish neural cells at different differentiation stages was further validated through transcriptomic analysis.
[1] Kestel Dévora. Draft intersectoral global action plan on epilepsy and other neurological disorders 2022–2031[EB/OL]. https://www.who.int/news/item/28-04-2022-draft-intersectoral-global-action-plan-on-epilepsy-and-other-neurological-disorders-2022-2031,2022-04-28/2024-07-10. [2] 贾建平, 陈生弟. 神经病学[M]. 第7版. 北京: 人民卫生出版社, 2013. [3] Chen Y, Xiong M, Dong Y, et al. Chemical control of grafted human psc-derived neurons in a mouse model of Parkinson’s disease[J]. Cell Stem Cell, 2016, 18(6): 817-826. [4] Grade S, Götz M. Neuronal replacement therapy: previous achievements and challenges ahead[J]. npj Regenerative Medicine, 2017, 2(1): 1-11. [5] Park SY, Park J, Sim SH, et al. Enhanced differentiation of human neural stem cells into neurons on graphene[J]. Advanced Materials, 2011, 36(23): H263-H267. [6] Liang P, Liu J, Xiong J, et al. Neural stem cell-conditioned medium protects neurons and promotes propriospinal neurons relay neural circuit reconnection after spinal cord injury[J]. Cell Transplantation, 2014, 23(Supp l): 45-56. [7] Brüstle O, Jones KN, Learish RD, et al. Embryonic stem cell-derived glial precursors: a source of myelinating transplants[J]. Science, 1999, 285(5428): 754-756. [8] Young JS, Kim JW, Ahmed AU, et al. Therapeutic cell carriers: a potential road to cure glioma[J]. Expert Review of Neurotherapeutics, 2014, 14(6): 651-660. [9] Martinez-Serrano A, Lundberg C, Horellou P, et al. CNS-derived neural progenitor cells for gene transfer of nerve growth factor to the adult rat brain: complete rescue of axotomized cholinergic neurons after transplantation into the septum[J]. Journal of Neuroscience, 1995, 15(8): 5668-5680. [10] Trujillo CA, Schwindt TT, Martins AH,et al. Novel perspectives of neural stem cell differentiation: from neurotransmitters to therapeutics[J]. Cytometry Part A: The Journal of the International Society for Analytical Cytology, 2009, 75(1): 38-53. [11] Denham M, Dottori M. Signals involved in neural differentiation of human embryonic stem cells[J]. Neurosignals, 2009, 17(4): 234-241. [12] Das KP, Freudenrich TM, Mundy WR. Assessment of pc12 cell differentiation and neurite growth: a comparison of morphological and neurochemical measures[J]. Neurotoxicology and Teratology, 2004, 26(3): 397-406. [13] Greene LA, Tischler AS. Establishment of a noradrenergic clonal line of rat adrenal pheochromocytoma cells which respond to nerve growth factor.[J]. Proceedings of the National Academy of Sciences, 1976, 73(7): 2424-2428. [14] Wiatrak B, Kubis-Kubiak A, Piwowar A, et al. PC12 cell line: cell types, coating of culture vessels, differentiation and other culture conditions[J]. Cells, 2020, 9(4): 958. [15] Haftbaradaran Esfahani P, ElBeck Z, Sagasser S, et al. Cell shape determines gene expression: cardiomyocyte morphotypic transcriptomes[J]. Basic Research in Cardiology, 2019, 115(1): 7. [16] Feldman D, Singh A, Schmid-Burgk JL, et al. Optical pooled screens in human cells[J]. Cell, 2019, 179(3): 787-799. [17] Perlman ZE, Slack MD, Feng Y, et al. Multidimensional drug profiling by automated microscopy[J]. Science, 2004, 306(5699): 1194-1198. [18] Yuan SH, Martin J, Elia J, et al. Cell-surface marker signatures for the isolation of neural stem cells, glia and neurons derived from human pluripotent stem cells[J]. PLoS ONE, 2011, 6(3): e17540. [19] Ilieva M, Della Vedova P, Hansen O, et al. Tracking neuronal marker expression inside living differentiating cells using molecular beacons[J]. Frontiers in Cellular Neuroscience, 2013, 7: 266. [20] Vieira MS, Santos AK, Vasconcellos R, et al. Neural stem cell differentiation into mature neurons: mechanisms of regulation and biotechnological applications[J]. Biotechnology Advances, 2018, 36(7): 1946-1970. [21] Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks[J]. Commun. ACM, 2017, 60(6): 84-90. [22] Zhou J, Theesfeld CL, Yao K, et al. Deep learning sequence-based ab initio prediction of variant effects on expression and disease risk[J]. Nature Genetics, 2018, 50(8): 1171-1179. [23] Alipanahi B, Delong A, Weirauch MT, et al. Predicting the sequence specificities of dna- and RNA-binding proteins by deep learning[J]. Nature Biotechnology, 2015, 33(8): 831-838. [24] Singhal A, Phogat M, Kumar D, et al. Study of deep learning techniques for medical image analysis: a review[J]. Materials Today: Proceedings, 2022, 56: 209-214. [25] Saint M, Bertaux F, Tang W, et al. Single-cell imaging and RNA sequencing reveal patterns of gene expression heterogeneity during fission yeast growth and adaptation[J]. Nature Microbiology, 2019, 4(3): 480-491. [26] Yosinski J, Clune J, Bengio Y, et al. How transferable are features in deep neural networks?[C]//Advances in Neural Information Processing Systems 27. Montreal: Neural Information Processing Systems Foundation Inc, 2014: 3320-3328. [27] Pawlowski N, Caicedo JC, Singh S, et al. Automating morphological profiling with generic deep convolutional networks[EB/OL]. https://www.biorxiv.org/content/10.1101/085118v1,2016-11-02/2024-08-20. [28] Buggenthin F, Buettner F, Hoppe PS, et al. Prospective identification of hematopoietic lineage choice by deep learning[J]. Nature Methods, 2017, 14(4): 403-406. [29] Kraus OZ, Ba JL, Frey BJ. Classifying and segmenting microscopy images with deep multiple instance learning[J]. Bioinformatics, 2016, 32(12): i52-i59. [30] Nitta N, Sugimura T, Isozaki A, et al. Intelligent image-activated cell sorting[J]. Cell, 2018, 175(1): 266-276. [31] Crissman HA, Hirons GT. Methods in Cell Biology [M]. San Diego: Academic Press, 1994: 195-209. [32] Alford R, Simpson HM, Duberman J, et al. Toxicity of organic fluorophores used in molecular imaging: literature review[J]. Molecular Imaging, 2009, 8(6): 7290.2009.00031. [33] Yao K, Rochman ND, Sun SX. Cell type classification and unsupervised morphological phenotyping from low-resolution images using deep learning[J]. Scientific Reports, 2019, 9(1): 13467. [34] He T, Zhang Z, Zhang H, et al. Bag of tricks for image classification with convolutional neural networks[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach: Computer Vision Foundation, 2019: 558-567. [35] Kang K, Pang G, Zhao X, et al. A new benchmark for instance-level image classification[J]. IEEE Access, 2020, 8: 70306-70315. [36] Sokolova M, Japkowicz N, Szpakowicz S. Beyond accuracy, F-score and ROC: a family of discriminant measures for performance evaluation[C]//AI 2006: Advances in Artificial Intelligence. Heidelberg: Springer Berlin, 2006: 1015-1021. [37] Flach P, Kull M. Precision-recall-gain curves: PR analysis done right[C]//Advances in Neural Information Processing Systems 28. Montreal: Neural Information Processing Systems Foundation Inc, 2015: 838-846. [38] Ge SX, Son EW, Yao R. IDEP: an integrated web application for differential expression and pathway analysis of RNA-seq data[J]. BMC Bioinformatics, 2018, 19(1): 534. [39] Ashburner M, Ball CA, Blake JA, et al. Gene ontology: tool for the unification of biology[J]. Nature Genetics, 2000, 25(1): 25-29. [40] Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with deseq2[J]. Genome Biology, 2014, 15(12): 550. [41] Mani S, Shen Y, Schaefer J, et al. Failure to express gap-43 during neurogenesis affects cell cycle regulation and differentiation of neural precursors and stimulates apoptosis of neurons[J]. Molecular and Cellular Neuroscience, 2001, 17(1): 54-66. [42] Dungrawala H, Bhat KP, Le Meur R, et al. RADX promotes genome stability and modulates chemosensitivity by regulating rad51 at replication forks[J]. Molecular Cell, 2017, 67(3): 374-386. [43] Marceau A H. Single-stranded DNA binding proteins: methods and protocols[M]. Totowa: Humana Press, 2012: 1-21. [44] Ngo L, Haas M, Qu Z, et al. TUBB5 and its disease-associated mutations influence the terminal differentiation and dendritic spine densities of cerebral cortical neurons[J]. Human Molecular Genetics, 2014, 23(19): 5147-5158. [45] Teufel L, Tummler K, Flöttmann M, et al. A transcriptome-wide analysis deciphers distinct roles of g1 cyclins in temporal organization of the yeast cell cycle[J]. Scientific Reports, 2019, 9(1): 3343. [46] Rizvi AH, Camara PG, Kandror EK, et al. Single-cell topological RNA-seq analysis reveals insights into cellular differentiation and development[J]. Nature Biotechnology, 2017, 35(6): 551-560. [47] Niu Z, Zhong G, Yu H. A review on the attention mechanism of deep learning[J]. Neurocomputing, 2021, 452: 48-62. [48] Selvaraju R, Cogswell M, Das A, et al. Grad-CAM: visual explanations from deep networks via gradient-based localization[C]//Proceedings of the IEEE International Conference on Computer Vision. Venice: Computer Vision Foundation, 2017: 618-626.
