Silicon-Based Surface-Enhanced Raman Scattering Sensor For Quantitative ATP Detection With Ratiometric Strategy And Artificial Intelligence DNA Analysis

Nucleic acid is a type of biological macromolecule composed of different nucleotides.In addition to being a carrier of genetic information,the nucleic acid with specific sequence has other specific function.Aptamers are a class of nucleic acid sequences that have been screened for specific binding to target molecule.For example,adenosine triphosphate(adenosine triphosphate,ATP)-specific aptamer can be obtained by combing a single-stranded DNA library with ATP.On the one hand,specific detection can be achieved by utilizing the specific function of nucleic acid.On the other hand,the detection and sequence analysis of nucleic acid has important genomics implications.Therefore,researchers need to develop technology for rapid and accurate detection of nucleic acid.Surface-enhanced Raman scattering(SERS)technology can provide exponentially magnified Raman signal with excellent sensitivity and specificity,which has been widely used in the field of analysis and detection.Silicon nanohybrids-based SERS substrates(e.g.,two-dimensional silver nanoparticle modified silicon wafer(Ag NPs@Si)and three-dimensional gold nanoparticle modified silicon nanowire arrays(Au NPs@Si NWAr))simultanuously have significant SERS enhancement and excellent SERS signal reproducibility.On the other hand,the development of artificial intelligence requires high-quality database as the basis for machine learning.SERS technology can provide sensitive Raman signal containing rich molecular fingerprint information,and is therefore expected to be useful for building high-quality database for artificial intelligence.The work of this paper focuses on silicon-based nanohybrids SERS substrate and nucleic acid analysis.Typically,one is to construct aptamer-based ratiometric SERS sensors for quantitative detection,and another is to construct nucleic acid-based SERS database for label-free artificial intelligence detection of nucleic acid.The details of the study are as follows:ATP Detection with Ratiometric Strategy and Artificial Intelligence DNA AnalysisIn the first part,silicon-based ratiometric SERS sensor is constructed by modifying functionalized double-stranded DNA structure on the silicon-based SERS substrates.The double-stranded DNA structure consists of ATP-specific aptamer and its complementary strand.The complementary strand is split into two discrete fragments: DNA-C1 and DNA-C2.Two different kinds of dye molecules(Cy3 and ROX)are respectively modified in the terminal of these two fragments.Cy3 provides the internal standard signal and ROX provides the detection signal.DNA-C1 is covalently modified on the SERS substates by thiol.In the presence of ATP,ATP binds with aptamer to form a complex,leading to the dissociation of DNA-C2 from the double-stranded system.ROX thus leave away from the surface of substrates,thereby reducing the detection signal.Next,by hybridizing the new-added aptamer strand,the rigid double-stranded DNA structure is reconstructed.The distance between Cy3 and the surface of substrate restores to the previous state so that the internal standard signal remains unchanged.The resultant ratiometric signal(ie,the ratio of the detection signal to the internal standard signal)can be used for the quantitative detection of ATP with a detection limit of 9.12 p M.The detection results obtained through the slicon-based ratiometric SERS sensor are consistent with those of the commercial ATP detection kit.In addition,the sensor has good specificity and can be recyclablely utilized.In the second part,silicon-based SERS substrate and deep neural network(DNN)are combined to develop an artificial intelligence(AI)label-free DNA sensing strategy based on SERS database.The DNN model consists of input layer,multiple hidden layers,and output layer.First,oncogenes with different base lengths(p16-15 bp,p16-30 bp,p16-50 bp,p21-15 bp,p21-30 bp,p21-50 bp,p53-15 bp,p53-30 bp and p53-50 bp)is used as a model analyte and incubated directly with silicon-based SERS substrate.A large number of SERS spectra are collected and used to construct SERS database.Next,characteristic parameters(such as Raman peak number,Raman peak intensity,etc.)are extracted from the collected high-standard SERS spectra as input data for the DNN input layer.Input data are divided into training groups and validation groups according to a certain percentage.In the constructed DNN multi-layer,the non-linear data relationship of the training group data is obtained by back propagation(BP)algorithm to train the machine.After training,data of validation groups is used for the test of recognition rate.After training and testing,the accuracy rate of single target DNA is 86.11%,and the accuracy rate of binary and ternary mixed DNA samples is 65.63%.To be summarized,in this dissertation,the silicon-based ratiometric SERS sensor is constructed by modifying double-stranded aptamer structure for quantitative detection of ATP in sensitive and specific manners.On the other hand,artificial intelligence label-free DNA sensing strategy is developed by the combination of silicon-based SERS substrate and deep neural network.