用于将稳健检测限拟合到连续稀释数据的matlab软件.zip

# detection-limit-fitting Software for fitting robust detection limits (LODs) to serial dilution data. Fits can then be plotted individually or multiple on the same axes, and compared using t tests. [](https://uk.mathworks.com/matlabcentral/fileexchange/108694-detection-limit-fitting-tool) ## Discussions If you have feature requests or questions, or would like to get help with problem or report a bug, please use the discussions tab. ## Latest Revisions 1. **NEW FEATURE: ANOVA testing** 2. Changed t test to just be between two files, now that ANOVA is implemented. ## Installation ### Matlab app install (multiplatform) If you use Matlab and want to use the app inside Matlab, download build/LOD_calculations_beta_1.mlappinstall. Open with Matlab and it will install in My Apps with the name "LOD_calculations_beta_1". It can then be found and opened from the APPS tab in Matlab. Alternatively, open Matlab and click Add-Ons. Search for "Detection Limit Fitting Tool" and add to Matlab. ### Windows standalone installation If you want to use it as a standalone application in Windows, download build/LOD_calculations_beta_1.exe. Open and run and it will install as a standalone application. ### Mac standalone installation Coming soon... ## Usage The application has three tabs: 1. LOD calculation: for importing dilution series data and fitting to a choice of four models. Fits, data and plots can then be exported in various formats. 2. Multiple plots: for plotting multiple datasets and fits on the same axes. 3. T tests: for performing statistical comparisons of LODs. Note: it is possible to compare multiple datasets to a single dataset by individual t tests, but this is only robust when comparing multiple independent datasets to a single control dataset. ### Data format Data must be in .xlsx files with two columns. Column 1 must be concentration values and column 2 must be the corresponding signal values. Column headings are optional (see example files in /test), described below in 'Test Files' section. ### LOD calculation <img src="screenshots/LOD_calcs.png" width="500"> Click 'File browse' to select a file. There are then multiple parameters to select: 1. Model: chose between the four models. See https://doi.org/10.1016/j.bios.2022.114133 for descriptions of the models. 2. Concentration units: input the units of your analyte concentrations. 3. LOD False Negative Rate (%): input the value for the confidence level of a positive sample at the LOD (lower value gives higher LOD and a higher probability that a negative result is a real negative). 4. Blank False Positive Rate (%): input the value for the confidence level of the blank cutoff (lower value gives a high blank cutoff and higher LOD, giving a higher probability that a positive result is a real positive). 5. Variance Outlier Confidence Level (%): when finding the 'characteristic variance' to calculate the cutoff, a hypothesis test (G test) is performed to exclude outlier variances. This is the confidence level of that hypothesis test. See https://doi.org/10.1016/j.bios.2022.114133 for more details. 6. Confidence Level for LOD Interval (%): confidence level for the confidence interval of the LOD. <img src="screenshots/LOD_calcs_labelled.png" width="547"> Then click 'Fit' The data and fitted model should be plotted in the box on the left, and the LOD and fit statistics in the box underneath. <img src="screenshots/LOD_calcs_fitted.png" width="500"> If the fit is unsuccessful, try a different model. If still unsuccessful, please use the forum to discuss. Save Figure: to save the plot in various formats Save Data: Export a .xlsx file of the data, LOD, fitted parameters and statistics. Save .mat: Save data, LOD, fitted parameters and statistics in a .mat file. This is required for Multiple plots and T tests functionality. ### Multiple plots <img src="screenshots/multiple_1.png" width="500"> Click 'Add files' to add multiple files (either at once or can be added sequentially). The files must be the .mat files saved previously in the 'LOD calculation' tab. <img src="screenshots/multiple_2.png" width="500"> When all the files are loaded, use Ctrl/Shift and click to select which files to plot. **NOTE: all files must have the same units.** If you want to adjust the units, for example to convert from fM to M, you should fill 'Order of magnitude adjustment' with the number of orders of magnitude between your old and new units, taking account of the positive/negative sign (i.e. -15 for fM -> M). Also, fill in the 'New units'. Then select 'Colours' to choose the colourscheme. They are ColorBrewer schemes (can be browsed here: https://colorbrewer2.org/) Click 'Plot selected files' First without adjusting the units (leaving those boxes), so in fM: <img src="screenshots/multiple_3.png" width="500"> Secondly with a unit adjustment from fM to M: <img src="screenshots/multiple_4.png" width="500"> Save or clear plot using the buttons on the bottom right. ### T tests Two LODs can be compared using a t test: <img src="screenshots/ttest_1.png" width="500"> Select the two files to be compared with the 'Select file 1' and 'Select file 2' buttons. The file must be a .mat file saved previously in the 'LOD calculation' tab. <img src="screenshots/ttest_2.png" width="500"> **NOTE: files must have the same units.** Set the confidence level for the LOD confidence intervals. As previously, select which files to compare and click 'Compare'. Note: each file must have a unique name otherwise it will not run. The results appear below: <img src="screenshots/ttest_3.png" width="500"> ### ANOVA When comparing more than two LODs, ANOVA must be used to prevent type 1 errors (false-positive significance). There is now a tab for performing ANOVA testing: <img src="screenshots/anova1.png" width="500"> First, select the files to compare using the 'Select files' button: <img src="screenshots/anova2.png" width="500"> **NOTE: files must have the same units.** ANOVA evaluated the null hypothesis that all LOD values are equal, and gives a corresponding single p-value for all LODs. Post-hoc comparisons use the results of the ANOVA to tell you which LODs are significantly different from whcih other means. To perform, set the confidence level for the LOD post-hoc tests. Then choose which post-hoc test to use: - Tukey-Kramer: compares each LOD to all the other LODs, so best used when there is no heirarchy - Dunnett: compares all LODs to a signle reference LOD, so best used when there is an existing reference/gold-standard assay. The first dataset in the file list is used as the reference dataset, so make sure to select it on its own first with the 'Select files' dialogue, click open, then repeat selecting all the other files (singly or multiply). Click 'Compare' to perform the ANOVA: <img src="screenshots/anova3.png" width="500"> The one-way ANOVA results are diplayed in the box on the left, and the post-hoc testing is displayed in the table on the right. The difference of the log LODs is displayed along with confidence intervals at the confidence level set previously and the p-value of each comparison. If the confidence interval includes 0, the difference is not significant at that level. Depending on the size of your window, you may have to scroll left to see the whole table: <img src="screenshots/anova4.png" width="500"> ## Test files ### Linear model fit Use test_file_1.xlsx. Use model "Linear", and concentration units "fM" Data from: Miller, B.S., Bezinge, L., Gliddon, H.D. et al. Spin-enhanced nanodiamond biosensing for ultrasensitive diagnostics. Nature 587, 588–593 (2020). https://doi.org/10.1038/s41586-020-2917-1 ### Langmuir, 4PL and 5PL fits Use test_file_2.xlsx and test_file_3.xlsx. Use model "Langmuir", "4PL" or "5PL" and concentration units "fM" Data from: Benjam