| Methicillin-resistant staphylococcus aureus(MRSA)is one of the most common clinical pathogens in the burn ward.It has a special resistance mechanism,making the ?-lactam antibiotics and its complex inhibitor failure.The bacteria has a stronger pathogenic than ordinary staph and can cause serious complications such as sepsis and septic shock.However,there are some disadvantages such as time-consuming and cumbersome operations in traditional detection methods which are commonly used in clinical detection.The situation is not conducive to reduce the occurrence of the abuse of antibiotics.There are also some disadvantages such as high cost and professional requirements for technical personnel in the new detection method based on resistance gene.As a result,it hinder the promotion of the new technology because of the disadvantages.It has an important significance to find a rapid,accurate,convenient and economical detection method for clinical anti-infection treatment with MRSA.The near infrared spectral analysis technology is one of the fastest growing and the most striking analysis technology.It reflects the absorption spectra mainly about the overtone band and combination band because of the molecular vibrational in the organic compounds.So it can gain the characteristic information by analyzing the near infrared spectrum in the substances containing hydrogen group.Near infrared spectroscopy analysis method has great potentials such as time saving,convenience,high accuracy and lower cost,besides,it does not destroy the sample and pollute the environment by no consumption of chemical reagents.It meets our requirements for the detection of drug-resistant bacteria.Some multivariate chemometrics algorithms will be introduced for data processing and modeling.Therefore,we explore the feasibility which to identify methicillin-resistant staphylococcus aureus(MRSA)and methicillin-sensitive staphylococcus aureus(MSSA)by using the near infrared spectra(NIR)analysis technology combining with the support vector machine(SVM)algorithm in order to establish a new method for detection the drug-resistant bacteria.Methods:1.At first,we will make the bacteria MRSA(ATCC43300)and MSSA(ATCC25923)activated,separated,purified.And then,pick a single colony and vaccinate it to a test tube containing 5 ml LB fluid medium.Bacteria will be expanded for 18 hours in a constant temperature oscillation box in order to be made as the seed liquid.Then,200 ul seed liquid will be added to a test tube containing 5 ml LB fluid medium to expand for 24 hours.During this time,we test the optical density(OD)value of the bacteria solution every two hours.We take the time as abscissa and optical density value as vertical to draw a curve f or bacterial growth and confirm the logarithmic phase for S.aureus bacteria.2.Bacteria solution on logarithmic phase will be diluted into 7 solutions with different concentration and measured the optical density value at the same time.Multiple proportion dilution which has a 10 times dilution rate will be made more than 7 times for the diluted bacteria solution above.Plate count method is used to calculate the concentration for the original bacteria solution.And then,to establish the relationship curve between the actual concentration and the optical density value.3.Divide research subjects into group MRSA and group MSSA scientifically,with the number of 24 bacteria samples in each group.Every bacteria sample is expanded to the state on logarithmic phase and to the bacteria solution with the standard concentration of 1 x 109 cfu/ml.4.Connect the light source,the near infrared spectrometer,optical fiber sensor and computer,and then open the spectral acquisition software called spectrasuite.Debug signals and set parameters.we collect the spectrum signal of samples when the state of spectrometer is stable.We take the spectrum signal of each bacteria sample 20 times clockwise and gather a total of 960 spectra samples.5.Spectral data procured by Spectrasuite software will be exported.We take the wavelength as X axis and the reflectivity as the Y axis to make an original spectra curve of the MRSA and the MSSA respectively.In order to know the reproducibility of each spectral data in group MRSA and group MSSA quantitatively,we use the discrimination index D value for reference.6.We take the original spectral curve coming from two kinds of bacteria some preprocessing such as smoothing denoising,baseline correction,the first order derivative,making peak and so on,in order to make the difference between the spectra more significant.Attribution analysis and correlation analysis of the absorption peak b and will be executed to find out the difference and the meaning of difference between the MRSA and the MSSA.7.Principal component analysis(PCA)is taken to deal with the original spectral data with the hope of compress the number of variable.We use the new variables to make the 3 d scatter plot and analysis the result of classification through projection clustering distribution.The scores of the component compressed by the principal component analysis are taken as the input vector for support vector machine(SVM)to establish an identification model between the MRSA and the MSSA.At last,we calculate the classification accuracy and prediction accuracy with three kinds of kernel function which are linear,polynomial,radial basis function(RBF)model with 2/3 samples as training set,the remaining one-third of the sample as the prediction set.Before the identification model was established,we need to verify the best kernel parameters for radial basis kernel function and polynomial kernel function.8.We use disc diffusion method to identify clinical isolates whether it is the MRSA bacteria.The appraisal results will be used as a reference.According to preliminary method which to construct the identification model between the MRSA and MSSA,we detect the clinical isolates using near infrared spectroscopy.9.The veracity for the new method using NIR will be assessed based on the method in disc diffusion method and after that we will compare the method by NIR with other methods from many aspects such as experiment time,accuracy,operation steps,the price and so on.Results:1.We find that the optical density value of the bacteria grow rapidly during the time 8 to 14 hours which correspond to the bacteria logarithmic phase by analyzing the growth curve of the MRSA and the MSSA.The growth trend of the two kinds of bacteria seems to be similar.We choose the 14 h as the time which amplify the bacteria for the follow-up study.2.According to the relationship between the optical density value and the actual concentration of the bacterial solution in the logarithmic phase,we calculate the formula of the concentration curve,for example,MSSA: y = 4.841 x-0.053(R2 = 0.9931);MRSA: y = 5.466 x-0.049(R2 = 0.9924).The unit is 109 cfu/ml.We decide to choose 1×109 cfu/ml as the experimental strain detection concentration.3.In the repeatability research of the MRSA and the MSSA spectral curve,we know that MRSA ranges from 0 to 2 with the mean value of 0 and MSSA ranges from 0 to 1 with the mean value of 0 according to the result of identification index Dy1y2.The repeatability in the research is fine and results are reliable.4.After the spectral preprocessing,we find the absorption peaks in 950.1 nm?1140.92 nm?1330.63 nm?1383.89 nm ?1494.81 nm?1748.28 nm?1857.1 nm?1927.59 nm and 2021.16 nm.We can not find the obvious difference between the MRSA and the MSSA by observing peak's position,waveform and peak value,even if deal with the absorption peak using the correlation coefficient analysis.5.The results which first three principal components occupy the total spectrum variability are 81.482%,6.247% and 3.077% respectively,and the accumulative contribution rate of the first three major factors is over 90%,maintaining most of the information for original characters.According to principal component cluster 3 d map based on first 3 major factors,it displays a good result to distinguish the MRSA from the MSSA.6.We use the training set to verify the optimum parameters for kernel paramete before set models by the support vector machine(SVM)and get the results that kernel parameters for polynomial is 8 and radial basis function is 1.1.The classification accuracy of the training set are as follows:the linear kernel function is 96.02% ? 0.50%;the polynomial kernel function is 98.73% ? 0.31%;the radial basis kernel function is 99.72% ? 0.21%.The prediction accuracy of the test set are as follows: the linear kernel function is 95.69%?0.01%;the polynomial kernel function is 98.88%?0.00%;the radial basis kernel function is 99.47%?0.00%.It preliminarily reflects that identification model for MRSA established by near infrared spectroscopy combining with support vector machine(SVM)algorithm has a high classification accuracy.7.When using identification model established by near infrared spectroscopy combining with support vector machine(SVM)algorithm to detect the S.aureus bacteria from clinical isolates,the classification accuracy of training set is 98.03% and the classification accuracy of test set is 98.5%.It further reflects that the near infrared spectral analysis technology has a very high accuracy to distinguish MRSA and MSSA in clinical isolates.8.Take the appraisal result obtained by gold standard disc diffusion method for reference,the same percentage of the NIR method is 98.5%,and the sensitivity is 97.00%,the specificity is 100.00%.The column contact number phi = 0.971,Kappa value is 0.970 after using pearson?2 inspection.It reflects that the NIR method has a strong positive correlation and a strong consistency compared with the disc diffusion method.9.Compare the different method for MRSA detection.On the accuracy,the result in near infrared method is 98.5% that is higher than those reported in literature suc h as MIC broth dilution method(91%),drug resistant protein assay(97.6%);on the time consuming,it takes 14 h for near infrared method less then the one costs by the disc diffusion method(24-48 h),MIC broth dilution method(24 to 36 h)and AGAR dilution method(24 h);on the operation,the near infrared method is simple.The technical personnel who have been trained simply can operate it independently;on the price,it just costs 10-20 yuan for consumables every times which is cheaper than the disc diffusion method(80 yuan/times)and broth dilution method(250 yuan/times),much cheaper than the PCR method and PBP2 a method.Conclusion:It is feasible to establish a identification model for MRSA and MSSA using near infrared spectroscopy combined with support vector machine(SVM)algorithm.Comparing to the other detection such as disc diffusion method,the minimum bacteriostasis concentration method,the PCR method and the resistant protein test in accuracy,time-consuming,operation and cost,the NIR detect method has a comprehensive advantage.With the characteristic of accurate,rapid,convenient,economic and so on,it is expected to become a new kind of detection method for drug-resistant bacteria and to be widely used in grassroots medical institutions.The new detect method will have great significance to control the infection and guide the clinical treatment. |
PDF Full Text Request