文献类型： 外文期刊

作者： Feng Xu-ping ¹ ; Peng Cheng ³ ; Zhang Chu ¹ ; Liu Xiao-dan ¹ ; Shen Ting-ting ¹ ; He Yong ¹ ; Xu Jun-feng ³ ;

作者机构： 1.Zhejiang Univ, Coll Biosyst Engn & Food Sci, Hangzhou 310058, Zhejiang, Peoples R China

2.Minist Agr, Key Lab Spect, Hangzhou 310058, Zhejiang, Peoples R China

3.Zhejiang Acad Agr Sci, Inst Qual & Stand Agroprod, Hangzhou 310021, Zhejiang, Peoples R China

关键词： NIR hyperspectral imaging;CRISPR/Cas9;Radial basis function neural network;Visualization

期刊名称：SPECTROSCOPY AND SPECTRAL ANALYSIS （ 影响因子：0.589； 五年影响因子：0.504 ）

ISSN： 1000-0593

年卷期： 2018 年 38 卷 2 期

页码：

收录情况： SCI

摘要： Mutant screening is an important step for CRISPR/Cas9 gene editing technology employed in crop breeding program. The present study proposes a visual identification method of CRISPR/Cas9-induced rice mutants based on near-infrared hyper-spectral image technology. A total of 1 200 samples of rice seeds were collected, comprising 600 wide types and 600 CRISPR/Cas9-induced mutant samples. The whole data set was divided into two groups according to the Kennard-Stone algorithm, a calibration set (400 samples) and a prediction set (200 samples) for each class. 24 optimal wavelengths were selected by 2nd spectra algorithm after preprocessing the selection spectral region with absolute noises by wavelet transform. Radial basis function neural network (RBFNN), extreme learning machine (ELM) and K-nearest neighbor (KNN) were used to build discrimination models based on the preprocessed full spectra and feature wavelengths. The results demonstrated that neural networks models achieved good recognition ability. The RBFNN model calculated on the optimal wavelength showed classification rates of 92.25% and 89.50% for calibration set and prediction set, respectively. Finally, the classification of mutant seeds could be visualized on prediction maps by predicting the features of each pixel on individual hyperspectral image based on 2nd derivative-RBFNN model. It was concluded that hyperspectral imaging together with chemometric data analysis was a promising technique to identify CRISPR/Cas9-induced rice mutants, which offered a powerful tool for evaluating large number of samples from CRISPR/Cas9 gene editing performance trials and breeding programs.