--- title: "NMRphasing" author: "Aixiang Jiang" date: "`r Sys.Date()`" output: rmarkdown::html_vignette vignette: > %\VignetteIndexEntry{NMRphasing} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- # NMRphasing NMRphasing is an R package designed for the correction of phase errors in NMR data. While its primary focus is on 1D NMR data, it can also be employed for 2D and 3D NMR data by processing one 1D NMR data file at a time. ## I. NMRphasing installation First of all, please ensure that you have the R package MassSpecWavelet, which is a dependency for NMRphasing. You can install MassSpecWavelet using the following example code: if (!requireNamespace("BiocManager", quietly = TRUE)) install.packages("BiocManager") BiocManager::install("MassSpecWavelet") NMRphasing is available on CRAN, and we can directly install it in R: ```{r eval = FALSE} install.packages("NMRphasing") ``` Alternatively, let's proceed to install NMRphasing from GitHub. ```{r eval = FALSE} devtools :: install_github(repo = "ajiangsfu/NMRphasing",force = TRUE) ## if you do not have old versions of NMRphasing, please remove force = TRUE ``` ## II. Data format The input data for NMRphasing can be in one of four formats: 1) A vector of absorption spectrum. 2) A complex vector. 3) A data matrix with two columns of spectrum data, where the first column represents the absorption spectrum, and the second column represents the dispersion spectrum. 4) A data frame with two columns of spectrum data, where the first column is for the absorption spectrum, and the second column is for the dispersion spectrum. After installing NMRphasing, you can load a subset of example data from our multiple metabolite spike-in experiment using the following code snippet: ```{r} library(NMRphasing) data("fdat") str(fdat) ``` In the code above, we load the dataset 'fdat' from the package. 'fdat$frequency_domain' represents the complex vector used in the following illustration, while 'freq' is utilized for plotting purposes. If your NMR data uses ppm values, you can substitute 'ppm' accordingly. ```{r dpi = 200} ## in order to make comparison, absorption part can be extracted fdat$Observed_Absorption = Re(fdat$frequency_domain) library(ggpubr) p1 = ggplot(fdat, aes(x = ppm, y = Observed_Absorption)) + geom_line() + theme_bw() + labs(y = "Observed Absorption") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) p1 ``` The observed absorption spectrum without phase error correction looks very bad. ## III. Example Code In this package, there are currently nine NMR phase error correction methods implemented. They are as follows: "NLS", "SPC_DANM","MPC_DANM", "SPC_EMP", "MPC_EMP","SPC_AAM","MPC_AAM", "SPC_DSM" and "MPC_DSM". ### 1. NLS This is our new shrinkage method, which is the fastest phase error correction method since it does not involve any optimization step. ```{r dpi=200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "NLS") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` The default setting for 'withBC' is TRUE, which tests for baseline bias based on spline regression on the lowess line. If the maximum of adjusted squared r is greater than 0.2, baseline correction is performed with modified polynomial fitting. If you set 'withBC' as FALSE, then no baseline bias will be tested and corrected. The example code is: ```{r eval=FALSE} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "NLS", withBC = FALSE) ``` ### 2. SPC_DANM The "SPC_DANM" method employs the traditional single linear model approach for phase error correction but incorporates a new optimization function designed to minimize the difference between absolute area and net area. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "SPC_DANM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 3. MPC_DANM This approach utilizes our new multiple linear model for phase error correction along with our optimization function designed to minimize the difference between absolute area and net area. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "MPC_DANM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 4. SPC_EMP This is an old phase error correction method based on the traditional single model approach targeting on an existing optimization function: entropy minimization with negative peak penalty. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "SPC_EMP") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 5. MPC_EMP This is to apply our new multiple linear model approach for phase error correction based on an existing optimization function: entropy minimization with negative peak penalty. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "MPC_EMP") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 6. SPC_AAM This is an old phase error correction method based on the traditional single model approach with an existing optimization function: absolute area minimization. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "SPC_AAM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 7. MPC_AAM This is to apply our new multiple linear model approach for phase error correction based on an existing optimization function:absolute area minimization. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "MPC_AAM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 8. SPC_DSM This is an old phase error correction method based on the traditional single model approach with an existing optimization function: summation minimization on dispersion. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "SPC_DSM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 9. MPC_DSM This is to apply our new multiple linear model approach for phase error correction based on an existing optimization function: summation minimization on dispersion. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "MPC_DSM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 10 SPC_ADSM This is a phase error correction method based on the traditional single model approach with an optimization function: absolute summation minimization on dispersion. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "SPC_ADSM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ### 11. MPC_ADSM This is to apply our new multiple linear model approach for phase error correction based on an optimization function: absolute summation minimization on dispersion. ```{r dpi = 200} fdat$Phased_Absoprtion = NMRphasing(specDatIn = fdat$frequency_domain, method = "MPC_ADSM") p2 = ggplot(fdat, aes(x = ppm, y = Phased_Absoprtion)) + geom_line() + theme_bw() + labs(y = "Phased Absoprtion") + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_blank(), axis.line = element_line(colour = "black"), axis.title.x = element_text(size = 6), axis.title.y = element_text(size = 6)) ggarrange(plotlist = list(p1,p2),labels = c("Before","After"),font.label = list(size = 6),nrow = 2, ncol=1) ``` ## IV. NMRphasing R package general information Version: 1.0.5 Authors@R: c( person("Aixiang", "Jiang", role = c("aut", "cre", "cph"), email = "aijiang@bccrc.ca", comment = c(ORCID = "0000-0002-6153-7595")) ) Maintainer: Aixiang Jiang Depends: R (>= 4.3.0),stats Suggests: knitr, rmarkdown, ggpubr VignetteBuilder: knitr Imports: baseline, splines, MassSpecWavelet Description: There are three distinct approaches for phase error correction, they are: a single linear model with a choice of optimization functions, multiple linear models with optimization function choices and a shrinkage-based method. The methodology is based on our new algorithms and various references: Binczyk et al. (2015) Chen et al. (2002) de Brouwer (2009) Džakula (2000) Ernst (1969) Liland et al. (2010) License: MIT + file LICENSE Encoding: UTF-8 LazyData: true RoxygenNote: 7.2.3