Reconstructing cell cycle and disease progression using deep learning
Reconstructing cell cycle and disease progression using deep learning
复制标题
使用深度学习重建细胞周期和疾病进展
DOI：
10.1101/081364
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Eulenberg P
中科院分区：
文献类型：
--
作者：
Eulenberg P
成像流式细胞术将流式细胞术的荧光灵敏度和高通量功能与单细胞成像相结合，从而提供与深度学习优势相匹配的大量数据。我们介绍了DeepFlow，这是一种用于成像流式细胞术的数据分析工作流程，它将深度卷积神经网络与非线性降维相结合。DeepFlow使用神经网络的学习功能来可视化，组织和生物解释单细胞数据。将细胞周期作为细胞间变异性的一个来源，对于定量单细胞生物学至关重要。我们为细胞循环Jurkat细胞的大型数据集演示了DeepFlow。首先，我们从原始图像数据中重建细胞在细胞周期中的连续进展。这表明DeepFlow可以学习分类表型之间的连续距离度量。其次，我们能够检测和分离死细胞的亚群，尽管数据集已经使用既定方法进行了清理。DeepFlow以无监督的方式检测这种形态异常的亚群。第三，在细胞周期阶段的无标记分类中，与最近基于一系列图像特征提升的方法相比，我们的错误率降低了6倍。与以前的方法相比，DeepFlow的预测速度足够快，可以考虑与成像流式细胞术测量过程集成。作者摘要DeepFlow是一种基于深度学习的数据分析工作流程，针对成像流式细胞术的要求进行了优化。我们用它来分析一个大型数据集的某种类型的人类T细胞（Jurkat细胞），其中经历细胞周期。DeepFlow能够重建这些细胞的连续细胞周期进程，并将死细胞与活细胞分离。我们展示了神经网络的学习特征如何被可视化和生物学解释。当用于对细胞周期阶段进行分类时，DeepFlow的表现明显优于以前的方法。
Imaging flow cytometry combines the fluorescence sensitivity and high-throughput capabilities of flow cytometry with single-cell imaging, and hence provides high-volume data well-matched to the strengths of deep learning. We present DeepFlow, a data analysis workflow for imaging flow cytometry that combines deep convolutional neural networks with non-linear dimension reduction. DeepFlow uses learned features of the neural network to visualize, organize and biologically interpret single-cell data. Dissecting the cell cycle as a source of cell-to-cell variability is crucial for quantitative single-cell biology. We demonstrate DeepFlow for a large dataset of cell-cycling Jurkat cells. First, we reconstruct the cells’ continuous progression through cell cycle from raw image data. This shows that DeepFlow can learn a continuous distance measure between categorical phenotypes. Second, we are able to detect and separate a subpopulation of dead cells, although the data set had been cleaned using established approaches. DeepFlow detects this morphologically abnormal subpopulation in an unsupervised manner. Third, in label-free classification of cell cycle phases, we reach a 6-fold reduction in error rate as compared to a recent approach based on boosting on a series of image features. In contrast to previous methods, DeepFlow’s predictions are fast enough to consider integration with the imaging flow cytometry measurement process.Author SummaryWe present DeepFlow, a deep learning based data analysis workflow optimized for the requirements of imaging flow cytometry. We use it to analyze a large data set of a certain type of human T cells (Jurkat cells), which undergo cell cycle. DeepFlow enables reconstructing the continuous cell cycle progression of these cells, and separates dead from living cells. We show how learned features of the neural network can be visualized and biologically interpreted. When used to classify the cell cycle stage, DeepFlow performs significantly better than previous approaches.