Virtual perturbations to assess explainability of deep-learning based cell fate predictors

Virtual perturbations to assess explainability of deep-learning based cell fate predictors
复制标题

虚拟扰动评估基于深度学习的细胞命运预测因子的可解释性

DOI:
10.1101/2023.07.17.548859
复制
发表时间:
2023
期刊:
--
影响因子:
--
通讯作者:
Soelistyo C
Soelistyo C
中科院分区:
--
文献类型:
--
作者:
Soelistyo C

文献摘要

参考文献

被引文献

相似文献

可解释的深度学习在从实验观察中提取科学见解方面有着重要的前景。这在生物成像领域尤其如此,该领域的原始数据往往非常庞大,但极其多变,很难研究。然而,在深度学习辅助科学发现方面,一个长期存在的挑战是,人工神经网络的工作原理往往难以解释。在这里,我们提出了一种简单的方法来研究训练好的神经网络的行为:虚拟扰动。通过对输入数据或其内部表示进行精确和系统的更改,我们能够在深度学习模型的输出中发现因果关系,进而在潜在现象本身中发现因果关系。作为一个范例,我们使用了最近描述的基于深度学习的细胞命运预测模型。在机械细胞竞争的实验模型中,我们训练网络来预测不太适合的细胞的命运。通过将虚拟扰动应用于训练的网络,我们发现了细胞环境和最终命运之间的因果关系。我们将这些与正在研究的生物系统的已知属性进行比较,以证明该模型忠实地捕获了见解。
Explainable deep learning holds significant promise in extracting scientific insights from experimental observations. This is especially so in the field of bio-imaging, where the raw data is often voluminous, yet extremely variable and difficult to study. However, one persistent challenge in deep learning assisted scientific discovery is that the workings of artificial neural networks are often difficult to interpret. Here we present a simple technique for investigating the behavior of trained neural networks: virtual perturbation. By making precise and systematic alterations to input data or internal representations thereof, we are able to discover causal relationships in the outputs of a deep learning model, and by extension, in the underlying phenomenon itself. As an exemplar, we use a recently described deep-learning based cell fate prediction model. We trained the network to predict the fate of less fit cells in an experimental model of mechanical cell competition. By applying virtual perturbation to the trained network, we discover causal relationships between a cell's environment and eventual fate. We compare these with known properties of the biological system under investigation to demonstrate that the model faithfully captures insights.
DOI: 10.1038/nature13317
发表时间: 2014-05-22
期刊: NATURE
影响因子: 64.8
作者:
Martins, Vera C.;Busch, Katrin;Rodewald, Hans-Reimer
通讯作者: Rodewald, Hans-Reimer
使用 Auto ML 提高黑色素瘤诊断系统的性能
DOI: --
发表时间: 2023
期刊:
影响因子: --
作者:
Inoue Manabu;Yoshimoto Takeshi;Tanaka Kanta;Koge Junpei;Shiozawa Masayuki;Nishii Tatsuya;Ohta Yasutoshi;Fukuda Tetsuya;Satow Tetsu;Kataoka Hiroharu;Yamagami Hiroshi;Ihara Masafumi;Koga Masatoshi;Mlynash Michael;Albers Gregory W.;Toyoda Kazunori;正木達也・北畠直人・飛塚丈輝・花崎和寿・張 維倫・永岡 隆
通讯作者: 正木達也・北畠直人・飛塚丈輝・花崎和寿・張 維倫・永岡 隆
用于活细胞成像中染色质形态分类的卷积神经网络。
DOI: 10.1007/978-1-0716-2221-6_3
发表时间: 2022
期刊: Methods in molecular biology (Clifton, N.J.)
影响因子: --
作者:
Ulicna K
通讯作者: Ulicna K
DOI: 10.1016/j.devcel.2013.04.013
发表时间: 2013-05-28
期刊: DEVELOPMENTAL CELL
影响因子: 11.8
作者:
Tamori, Yoichiro;Deng, Wu-Min
通讯作者: Deng, Wu-Min
DOI: 10.1117/1.jbo.27.7.076003
发表时间: 2022-07
影响因子: 3.5
作者:
Moallem, Golnaz;Pore, Adity A.;Gangadhar, Anirudh;Sari-Sarraf, Hamed;Vanapalli, Siva A.
通讯作者: Vanapalli, Siva A.