シンポジウム
57 計算技術が拓く神経障害の解析：細胞からヒトまで
57 Intelligent analysis for neuronal disorders: from cell to human
座長：吉田 祥子（豊橋技術科学大学）・諫田 泰成（国立医薬品食品衛生研究所） |
|---|
| 2022年7月2日 16:10～16:30 ラグナガーデンホテル 羽衣：西 第10会場
3S10e-01
深層学習を用いた発達神経毒性誘発動物の神経形態変化定量解析技術
Quantitative analysis of neural morphological changes in developmental neurotoxicity-induced animals using deep learning
*田村 和輝(1)
1. 浜松医科大学
*Kazuki Tamura(1)
1. Hamamatsu University School of Medicine
Keyword: deep learning, Purkinje cell, neural morphological changes
We report an approach to discriminate neural morphology affected by toxicity using deep learning quantitatively. The effects of fetal toxicity exposure on the developing nervous system are often evaluated using postnatal behavioral abnormalities as an indicator. Our group evaluated the effects of toxicity on neurodevelopment by visualizing the location and morphology of Purkinje cell bodies around the V/VI lobes using confocal fluorescence microscopy of antibody-stained fresh cerebellar slices. A common problem in morphological evaluation using fluorescence images is how to evaluate the acquired images. Although the position and morphology of cell bodies can be felt by anyone who looks at the images, there are many cell bodies in borderline. In particular, in weak toxicity, which is difficult to observe in behavioral abnormalities, cell body irregularities are slight, and the ratio of cell bodies in the borderline is large. This study developed an algorithm to discriminate cell bodies into regular and irregular cell bodies based on uniform criteria using deep learning for discriminating cell bodies. However, by taking advantage of deep learning, we have made it possible to detect irregular cells using a realistic number of training data. Training and discrimination methods by Faster -R-CNN, which detects the position of an arbitrary object (e.g., a car) from a photograph, and ResNet, which identifies objects (e.g., signs), were provided by Matlab (Mathworks Inc.). Since deep learning research is constantly evolving and it is difficult for an outsider to use the latest technology, we used a network of the previous generation. The preparation was to annotate eight images manually and keep an image for test data. We created a patch from the eight images to determine whether a cell body is regular or irregular. Each image contains many cell bodies, such as 100 to 200 patches of regular cell bodies, 1 to 10 patches of irregular cell bodies were created. Faster-R-CNN detected more than 95% of the cell bodies. The accuracy of cell body discrimination of test data is generally found over 90%. This study showed that deep learning in cell body classification could be applied by applying basic deep learning methods and does not necessarily require a large data set. Through the application of deep learning, we hope that morphological evaluation that removes the dependency of the evaluator will be widely used. | | 2022年7月2日 16:30～16:50 ラグナガーデンホテル 羽衣：西 第10会場
3S10e-02
Three-dimensional acoustic impedance mapping of cultured biological cells
*Naohiro Hozumi(1)
1. Toyohashi University of Technology
Keyword: Cultured cell, Acoustic impedance, Acoustic microscopy
Biological cells change their internal three-dimensional structure and internal distribution of elastic parameters by division, differentiation, or lesions. The visualization of these changes in living cells is of great interest in medicine and biology. The three-dimensional acoustic impedance microscopy proposed by the presenters is a three-dimensional, non-invasive, continuous, and simple observation system for obtaining the distribution of acoustic impedance inside cultured cells. The acoustic impedance dealt with here is an elastic parameter that correlates to the bulk modulus. A focused ultrasound pulse, its frequency range spreading from 100 to 500 MHz is transmitted through the culture film from the rear side, and the reflected signal from inside the cells are acquired by the same transducer. At the same time, the reflected signals from the interface between the film and the culture medium is acquired. These signals are compared and converted to reflection coefficient at each point in the cell, and appropriate signal processing is applied using the known acoustic properties of the film and culture medium, to convert them to the acoustic impedance distribution in the thickness direction of the cell. By laterally scanning the ultrasonic beam, the acoustic impedance distribution inside a cell with a thickness of about 2 μm is drawn. The spatial resolution is about 0.2 μm in the thickness direction and 3 to 4 μm in the plane direction. By aligning the optical and acoustic axes, optical and acoustic images of the same cell can be displayed. In the presentation, changes in time of the 3D shape and internal structure of cells after stimulation with medication will be demonstrated, using several kinds of cell samples. | | 2022年7月2日 16:50～17:10 ラグナガーデンホテル 羽衣：西 第10会場
3S10e-03
iPS細胞技術及び疾患モデルマウスを用いた精神疾患の分子病態解析
Modeling psychiatric disorders with iPS cell technology and disease mouse models
*中澤 敬信(1)
1. 東京農業大学
*Takanobu Nakazawa(1)
1. Tokyo Univ of Agriculture
Keyword: schizophrenia, iPS cell, disease mouse model
Schizophrenia, a highly prevalent disorder affecting approximately one in 100 people, is a multifactorial disease in which genetic and environmental factors are intricately intertwined. To date, a fundamental treatment for this disease is lacking, and some patients are not adequately treated with the major antipsychotic drugs. Since symptoms of the disease are diverse and may vary from patient to patient, stratification of the disease based on molecular pathogenesis is important to understand the disease and to develop patient-selective treatment strategies. Analyzing iPS cell-derived neurons from patients enables direct analysis of live neurons with the same genetic background as the patient. It may therefore be important to analyze iPS cell-derived differentiated neurons from patients with genetic and clinical data for stratifying the disease based on molecular pathogenesis. Furthermore, it may also be important to conduct neural circuit and behavioral level analyses using a disease mouse model by introducing the genetic mutations occurring in the patients. In these analyses, given the central nervous system is consisted of heterogeneous neurons, comprehensive data-driven research for elucidating the characteristics of each minor cell population is needed for understanding the mechanisms of the respective specific brain functions and molecular pathogenesis of the disease. Such integrative research will not only clarify the molecular pathogenesis of schizophrenia but also stratify the disease, and ultimately lead to the success of central nervous system drug development. Here I introduce our recent studies on the molecular and cellular pathogenesis of schizophrenia using iPS cells from patients and disease mouse models by introducing the genetic mutations occurring in the patients. | | 2022年7月2日 17:10～17:30 ラグナガーデンホテル 羽衣：西 第10会場
3S10e-04
実験と神経筋骨格モデルによる歩行の適応メカニズムの解明
Experimental and computational neuro-biomechanical modeling approaches for unravelling neural and biomechanical mechanisms of locomotor adaptation
*柳原 大(1)
1. 東京大学
*Dai Yanagihara(1)
1. The University of Tokyo
Keyword: locomotion, adaptation, neuromusculoskeltal model
Identifying neural network mechanisms and making causal linkages of neuronal activities to specific behaviors is at the core of neuroscience. A range of effective manipulations have been developed in behavioral neuroscience for investigating the functional roles of particular neural systems in specific behaviors, such as the use of Designer Receptors Exclusively Activated by Designer Drugs (DREADD)-based chemogenetic and/or optogenetic controls that selectively target particular neuronal circuits.
Animals and humans show rapid adaptations in locomotor behavior within diverse environments. Neural control of locomotion involves circuits at different levels of the nervous system. The spinal locomotor circuits generate the temporal pattern of limb movements during locomotion, while the brainstem structures initiate locomotion and control its speed. The cerebellum is thought to adjust and fine-tune interlimb coordination by modulating activity in descending pathways. On the other hand, locomotion is generated by dynamic interactions among neural systems, musculoskeletal systems, and the environment. The results from previous anatomical and physiological studies have enabled the construction of realistic neural models. By incorporating information from corresponding models of the musculoskeletal system in rodents, integrated neuromusculoskeltal models have been developed that can be used to conduct forward dynamic simulations. In this short talk, we will consider the sensory-motor integration involved in the modulation of stepping patterns such as walking and trotting with forward speed, the control of stepping over an obstacle during locomotion on a flat walkway, and the adaptive control of locomotion on a split-belt treadmill. The combination of state-of-the-art experimental and computational modeling-dynamic simulation approaches provides a new depth of understanding of the neural mechanisms involved in locomotor adaptation. | | 2022年7月2日 17:30～17:50 ラグナガーデンホテル 羽衣：西 第10会場
3S10e-05
HLMとWGCNAを用いた妊娠期ストレスの出生児への影響解析
Hierarchical linear modeling and Weighted gene co-expression network analysis to the Stress in Pregnancy study
*野村 容子(1)
1. ニューヨーク市立大学クイーンズ校
*Yoko Nomura(1)
1. Queens Collage, City University of New York
Keyword: Stress in Pregnancy, placenta, Weighted gene co-expression network analysis, Hierarchical linear modeling
The prenatal period represents a sensitive window to both external and internal environmental cues, which may dramatically impact fetal/child development, potentially extending throughout life. Stress in Pregnancy (SIP) Study is a known contributor to maladaptive neurobehavioral development of the offspring; however, the underlying molecular mechanisms linking SIP with childhood outcome remain largely unknown. In this study, using a hierarchical linear model (HLM), we found that in utero Sandy exposure was associated with greater clinical (anxiety, depression, and somatization) and lower adaptive behaviors (social skills and functional communication), and evaluated inter-relationships of placental transcriptome-wide changes at the module level with SIP and infant temperament to identify functional networks/pathways that link SIP and offspring's temperament in early childhood. Transcriptome-wide gene expression data were generated using RNA-seq from placenta samples collected in a multi-ethnic urban birth cohort in New York City (n = 129). Weighted gene co-expression network analysis (WGCNA) was used to characterize placental co-expression modules, which were then evaluated for their associations with SIP and infant temperament. WGCNA revealed 16 gene co-expression modules. One module, enriched for regulation of chromosome organization/gene expression, was positively associated with SIP and negatively associated with Regulatory Capacity (REG), a component of infant temperament. Two other modules, enriched for co-translational protein targeting and cell cycle regulation, respectively, displayed negative associations with SIP and positive associations with REG. A module enriched with oxidative phosphorylation/mitochondrial translation was positively associated with REG. These findings support the notion that the placenta provides a functional in utero link between SIP and infant temperament, possibly through transcriptional regulation of placental gene expression. | | | |
|
|
