Biology subjects

Tokuoka, Y.

Publications and source records attributed to Tokuoka, Y..

8 recordsLinked to original sources

bioRxiv2025

Mixing features of transcription factors and genes enables accurate prediction of gene regulation relationships for unknown transcription factors

Identifying regulatory relationships between transcription factors (TFs) and genes is essential to understand diverse biological phenomena related to gene expression. Recently, deep learning-based models to predict TFs that bind to genes from nucleotide sequences of the target genes have been developed, yet these models are trained to predict known TFs only. In this study we developed a deep learning model, GReNIMJA (Gene Regulatory Network Inference by Mixing and Jointing features of Amino acid and nucleotide sequences), to predict gene regulation even by unknown TFs. Our model is designed to mix the features of the TF amino acid sequences and nucleotide sequences of the target genes using a 2D LSTM architecture and to perform binary classification with the aim of determining the presence or absence of a regulatory relationship. The accuracy of our model in predicting regulatory relationships was 84.4% for known TFs (higher than those of conventional models) and 68.5% for unknown TFs; the latter is an unsolved task for conventional deep learning-based models. We expect our model to advance identification of unknown gene regulatory networks and contribute to the understanding of diverse biological phenomena.

systems biology↗

bioRxiv2024

Four-dimensional label-free live cell image segmentation for predicting live birth potential of mouse embryos

1Selection of high-quality embryos is critical in assisted reproductive technology (ART), but it relies on visual assessment by experts, and the birth rate remains low. We previously developed a deep learning method to predict the birth of mouse embryos by quantifying the morphological features of cell nuclei. This method involves cell nuclear segmentation on fluorescence microscopy images, but fluorescence labeling of nuclei is not feasible in medical applications. Here, we developed FL2-Net, a nuclear segmentation method for time-series three-dimensional bright-field microscopy images of mouse embryos without fluorescence labeling. FL2-Net outperformed existing state-of-the-art segmentation methods. We predicted the birth potential of mouse embryos from the nuclear features quantified by bright-field microscopy image segmentation. Birth prediction accuracy of FL2-Net (81.63%) exceeded those of other methods and experts (55.32%). We expect that FL2-Net, which can quantify nuclear features of embryos non-invasively and with high accuracy, might be useful in ART.

systems biology↗

bioRxiv2024

Deep learning-based automated prediction of mouse seminiferous tubule stage by using bright-field microscopy

Infertility is a global issue, with approximately 50% of cases attributed to defective spermatogenesis. For studies into spermatogenesis and spermatogenic dysfunction, evaluating the seminiferous tubule stage is essential. However, the current method of evaluation involves labor-intensive and time-consuming manual tasks such as staining, observation, and image analysis. Lack of reproducibility is also a problem owing to the subjective nature of visual evaluation by experts. In this study, we propose a deep learning-based method for automatically and objectively evaluating the seminiferous tubule stage. Our approach automatically predicts which of 12 seminiferous tubule stages is represented in bright-field microscopic images of mouse seminiferous tubules stained by hematoxylin-PAS. For training and validation of our model, we created a dataset of 1229 tissue images, each labeled with one of 12 distinct seminiferous tubule stages. The maximum prediction accuracy was 79.58% which rose to 98.33% with allowance for a prediction error of {+/-}1 stage. Remarkably, although the model was not explicitly trained on the patterns of transition between stages, it inferred characteristic structural patterns involved in the process of spermatogenesis. This method not only advances our understanding of spermatogenesis but also holds promise for improving the automated diagnosis of infertility.

systems biology↗

bioRxiv2024

Inference of gene regulatory networks for overcoming low performance in real-world data

The identification of gene regulatory networks is important for understanding the mechanisms of various biological phenomena. Many methods have been proposed to infer networks from time-series gene expression data obtained by high-throughput next-generation sequencings. Such methods can effectively infer gene regulatory networks for in silico data, but inferring the networks accurately from in vivo data remiains a challenge because of the large noise and low time sampling rate. Here, we proposed a novel unsupervised learning method, Multi-view attention Long-short term memory for Network inference (MaLoN). It can infer gene regulatory networks with temporal changes in gene regulation using the multi-view attention Long Short-term memory model. Using in vivo benchmark datasets in Saccharomyces cerevisiae and Escherichia coli, we showed that MaLoN can infer gene regulatory networks more accurately than existing methods. The ablated models indicated that the multi-view attention mechanism suppressed false positives. The order of activation of gene regulations inferred by MaLoN was consistent with existing knowledge.

systems biology↗

bioRxiv2024

Cell segmentation without annotation by unsupervised domain adaptation based on cooperative self-learning

Semantic cell segmentation from microscopic images is essential for the quantitative evaluation of cell morphology. Although supervised deep-learning-based models offer accurate segmentation, their performance degrades for unknown cell types. To address this problem, unsupervised domain adaptation methods based on adversarial training, self-training, or a combination of these approaches have been developed in recent years. These methods train the model using inference labels from the unknown domain as pseudo labels with reliability to resolve the discrepancy between the features of the unknown and known domains. However, conventional methods require a predefined threshold to calculate pseudo-labels reliability, leading to costly hyperparameter tuning. Here, we developed an unsupervised domain adaptation for semantic cell segmentation with cooperative self-learning (CULPICO: Cooperative Unsupervised Learning for PIxel-wise COloring) that does not require predefined threshold of the pseudo-labels reliability. The proposed method consists of two independent segmentation models and a mutual exchange mechanism of inference data. The models infer a label probability at each pixel and generate a pseudo-label as unsupervised learning. The pseudo-labels created by each model are mutually used as ground truth in the other model. Loss function is corrected by considering pixel-level discrepancies between the label probabilities inferred by the two models. The proposed method, despite being an unsupervised learning method, can segment efficiently the unknown cell types without labels with an accuracy comparable to supervised learning models. Our method, which could solve the performance degradation problem without constructing new datasets, is expected to accelerate life science by reducing the cost of extracting quantitative biological knowledge.

systems biology↗

bioRxiv2021

Deep learning-based algorithm for predicting the live birth potential of mouse embryos

In assisted reproductive technology (ART), embryos produced by in vitro fertilization (IVF) are graded according to their live birth potential, and high-grade embryos are preferentially transplanted. However, the rate of live birth following clinical ART remains low worldwide, suggesting that grading is inaccurate. One explanation is that grading is classically based on the characteristic shape of embryos at a limited number of developmental stages and does not consider the shape of embryos and intracellular structures, e.g., nuclei, at various stages important for normal embryogenesis. Therefore, here we developed a Normalized Multi-View Attention Network (NVAN) that directly predicts live birth potential from nuclear structural features in live-cell fluorescence images taken of mouse embryos across a wide range of stages. The classification accuracy of our method was 83.87%, which greatly exceeded that of existing machine-learning methods and that of visual inspection by embryo culture specialists. By visualizing the features that contributed most to the prediction of live birth potential, we found that the size and shape of the cell nucleus at the morula stage and at the time of cell division were important for live birth prediction. We anticipate that our method will help ART and developmental engineering as a new basic technology for IVF embryo selection.

developmental biology↗

bioRxiv2018

Predicting the future direction of cell movement with convolutional neural networks

Image-based deep learning systems, such as convolutional neural networks (CNNs), have recently been applied to cell classification, producing impressive results; however, application of CNNs has been confined to classification of the current cell state from the image. Here, we focused on cell movement where current and/or past cell shape can influence the future cell fate. We demonstrate that CNNs prospectively predicted the future direction of cell movement with high accuracy from a single image patch of a cell at a certain time. Furthermore, by visualizing the image features that were learned by the CNNs, we could identify morphological features, e.g., the protrusions and trailing edge that have been experimentally reported to determine the direction of cell movement. Our results indicate that CNNs have the potential to predict the future cell fate from current cell shape, and can be used to automatically identify those morphological features that influence future cell fate.

systems biology↗

bioRxiv2018

Development of Convolutional Neural Network Based Instance Segmentation Algorithm to Acquire Quantitative Criteria of Mouse Development

In embryology, analytical approaches such as segmentation are applied to acquiring quantitative criteria from time-series three-dimensional microscopic images. When used to segment cells or intracellular organelles, several current machine learning techniques outperform traditional image processing algorithms. However segmentation algorithms still have unsolved problems, especially in bioimage processing. The most critical issue is that the existing algorithms can perform only semantic segmentation, which distin-guishes whether a pixel is within an object (for example, nucleus) or not. In this study, we implemented a novel segmentation algorithm, based on deep learning, which segments each nucleus and adds different labels to the detected objects. This segmentation algorithm is called instance segmentation. Our instance segmentation algorithm, implemented as a neural network, which we named QCA Net, substantially out-performed 3D U-Net, which is the best semantic segmentation algorithm that uses deep learning. Using QCA Net, we quantified the nuclear number, volume, surface area, and center of gravity coordinates during the development of mouse embryos. In particular, QCA Net distinguished nuclei of embryonic cells from those of polar bodies formed in meiosis. We consider that QCA Net can greatly contribute to bioimage segmentation in embryology by generating quantitative criteria from segmented images.

systems biology↗