Preparation Of Liquid Crystal Type Aptamer Sensors For Tumor Markers And Polarized Light Image Classification Study

In recent years,the liquid crystal aptamer biosensor constructed with aptamers as molecular recognition elements and liquid crystal(LC)molecules as signal amplification and transducing elements provides a new way for tumor marker detection.However,only a few liquid crystal aptamer sensors can realize the quantitative detection of tumor markers at the body fluid level,and most of them still stay in the quantitative determination of spiked solutions.Therefore,improving the specific binding of targets in the sensor,reducing the adsorption of non-targets in the sensor and establishing an effective polarized image processing method are the key problems to be solved to improve the performance of the liquid crystal aptamer sensor and realize the detection of tumor markers in actual samples.In this paper,the glycoprotein tumor marker CA19-9 is taken as the research object,and the liquid crystal aptamer sensor based on aptamer core sequence is constructed.On the one hand,it retains the ability of aptamer targeting recognition and binding to CA19-9,on the other hand,it effectively avoids the influence of long-chain aptamers on the orientation of liquid crystal molecules,and realizes the preparation of liquid crystal aptamer sensor and quantitative detection of CA19-9.In order to further improve the accuracy of quantitative detection of serum samples CA19-9 by this sensor,this paper discusses the correlation between the characteristic information of polarized microscopic images and the concentration of CA19-9 by using machine learning and deep learning respectively,so as to realize the classification of serum samples CA19-9.The details are as follows:(1)Preparation of liquid crystal aptamer sensor and detection of CA19-9.The liquid crystal aptamer sensor includes two layers of substrates:molecular oriented substrate and functionalized substrate.The molecular oriented substrate induces the ordered arrangement of 5CB liquid crystal molecules by DMOAP,and the functionalized substrate uses DMOAP/APTES-GA-Apt to specifically recognize and combine CA19-9,which directly breaks the ordered arrangement of LC molecules in the sensor,changes the transmission direction of polarized light,and changes the polarized light microscopic image from dark to bright.Under the best experimental conditions,there is a good linear relationship between the logarithm of the concentration of the spiked buffer CA19-9 and the bright area coverage(Br)of the polarized microscope image,with the linear range of 5.0～1.0×10~3 U/m L and the detection limit of 0.1009 U/m L.However,the accuracy of the sensor in specifically identifying CA19-9 and blood sample CA19-9 needs to be further improved.(2)Classification of machine learning methods for polarized images.The machine learning method is introduced to explore the feasibility of classifying polarized images of CA19-9 liquid crystal sensor with clinical cancer threshold as the dividing line by machine learning method and judging interferers.Firstly,KNN and SVM,two traditional machine learning methods,are used to classify polarized images,and the evaluation indexes of these two models under different parameters are calculated,through which the classification performance of the models is evaluated.The results show that the KNN classification model based on Manhattan distance formula is more suitable for the classification of polarized images in this study,and the classification accuracy can reach93.25%,which realizes the classification of CA19-9 clinical threshold,but it cannot accurately distinguish CA19-9 from other detected substances.(3)Classification of convolutional neural network for polarized images.In order to make the classification model recognize some characteristic information that can’t be recognized by traditional machine learning methods,convolutional neural network is introduced to realize the automatic classification of polarized images,and the classification standard is whether the concentration of CA19-9 is higher than the clinical detection threshold(400 U/m L).In this paper,VGG16 and 2D-CNN convolutional neural networks are used to classify polarized images,and the network structure suitable for this study is selected by calculating the evaluation indexes of the two network models,and it is used in unknown sample detection experiments and interference discrimination experiments.The results show that the classification performance of 2D-CNN is better,and the correct rate of its classification and identification of unknown samples is 99.57%,and it can accurately identify non-CA19-9 interferers.This method can not only improve the accuracy of the detection results of CA19-9 liquid crystal sensor,but also simplify the experimental steps to avoid errors caused by artificially calculating the bright area coverage of polarized images.It can also solve the experimental interference caused by sensor-specific defects,improve the experimental efficiency and accuracy as a whole,not only realize the early diagnosis of cancer patients by detecting CA19-9,but also provide a useful reference for the further development of liquid crystal sensors in the future.In this paper,an aptamer liquid crystal sensor for detecting tumor markers was constructed by truncating long-chain aptamers and using aptamer core sequences.Experiments show that truncated aptamer will reduce the specific binding ability between aptamer and target substance,and reduce the anti-interference ability of the sensor.Convolutional neural network model based on the characteristic information of polarized microscope image of the sensor can accurately identify high-concentration detection substances,effectively classify detection substances from other interfering substances,and make up for the lack of specificity of the sensor itself.The experimental results have important reference value for the research of liquid crystal aptamer sensors for other tumors and even more target molecules.