exmedt: Exposome Mediation Module

Author: Mengyuan Ren (myren@bjmu.edu.cn)

Date: 2022-11-25

The exmedt package is designed for estimating and screening potential mediation pathways of indigenous biomarkers (i.e., mediator) between external environmental exposure and health outcome in a user friendly and efficient way. Especially, when the number and category of exposures and mediators are of high dimension. Please see the website (http://www.exposomex.cn/#/expomediation) for more information.

Users can install the package using the following code:

if (!requireNamespace("devtools", quietly = TRUE)){

	install.packages("devtools")

	devtools::install_github('ExposomeX/excros/exmedt',force = TRUE)

	devtools::install_github('ExposomeX/extidy/extidy',force = TRUE) #"extidy" package is optional if the data file has been well prepared. However, the it is recommended as users may need tidy the data to meet the modeling requirement, such as deleting varaibles with low variance, transforming data type, classifying variable into several level, etc.
}

library(exmedt)
library(extidy)

Tips:

Before using the package, a user defined physical output path (i.e., OutPath) is recommended. For example

 OutPath = "D:/test" #The default path is the current working directory of R. Users can use this code to set the preferred path.

For each step, the returned values can be named as users' like by following R language requirement.
All the PID must be the same with the one provided by InitMedt function, e.g., res$PID.

Example codes:

1. Initial Mediation module:

res <- InitMedt()
res$PID

2. Load data for mediation analyses

res2 <- LoadMedt(PID=res$PID,
	            UseExample = "example#1",
	            DataPath = NULL,
	            VocaPath = NULL)
res2$Expo$Data
res2$Expo$Voca

3 Set Output physical path (optional)

OutPath = "E:/Medt_test" #The default path is the current working directory of R.

4 Tidy procedures

res3 = TransImput(PID=res$PID,
                  Group="F",
                  Vars="all.x",
                  Method="lod")
res4 = TransType(PID=res$PID,
                 Vars="C1",
                 To="factor")
				 
res5 = TransScale(PID=res$PID,
                  Group="F",
                  Vars="all.x",
                  Method="normal")
res6 = TransScale(PID=res$PID,
                 Group="F",
                 Vars="all.m",
                 Method="normal")

res7 = TransDistr(PID=res$PID,
                  Vars="all.x",
                  Method="ln")
res8 = TransDistr(PID=res$PID,
                 Vars="all.m",
                 Method="ln")
				 
res9 = TransDummy(PID=res$PID,
                   Vars="default")

4. Divide exposures and mediators into separate groups

res10 <- XMlists(PID=res$PID)
res10$ExpoList
res10$MediList

5. Pairwise mediation modelling

res11 <- Pairwise(PID=res$PID,
                  VarsY = "Y1",
                  VarsX = "X1,X2,X3",
                  VarsM = "M1,M2,M3",
                  VarsC = "C1,C2,C3",
                  Family = "linear",
                  Iter = 500)
res11$MedtPairWise_Stats
res11$MedtPairWise_Stats

6. Visualize pairwise mediation modelling result

res12 <- VizMedtPair(PID=res$PID,
                     Brightness = "light",
                     Pallette = "default1")
res12$plotdata
res12$plot

7. Exposures dimension reduction

res13 <- RedX(PID=res$PID,
                VarsY = "Y1",
                VarsC = "C1,C2,C3",
                VarsM = "M1,M2,M3",
                Method = "mean",
                Folds = 10,
                Family = "linear",
                Iter = 500)
res13$MedtRedX_ERSall
res13$MedtRedX_ERSlist
res13$MedtRedX_Stats

8. Mediators dimension reduction

res14 <- RedM(PID=res$PID,
              VarsY = "Y1",
              VarsX = "X1,X2,X3",
              VarsC = "C1,C2,C3",
              Method = "mean",
              Family = "linear",
              Iter = 500)
res14$MedtRedM_all
res14$MedtRedM_list

9. Exposure and mediator dimension reduction

res15 <- RedXM(PID=res$PID,
                 VarsY = "Y1",
                 VarsC = "C1,C2,C3",
                 Method = "mean",
                 Family = "linear",
                 Iter = 500)
res15$MedtRedXM_all
res15$MedtRedXM_list

10. Visualize MedtRedXM mediation modelling result

res16 <- VizRedXM(PID=res$PID,
                    Brightness = "light",
                    Pallette = "nature")
res16$MedtRedXM_plotadta
res16$MedtRedXM_plot

11. Estimation the importance of mediators

res17  <- ImptM(PID = res$PID,
                 VarsY = "Y1",
                 VarsX = "X1,X2,X3",
                 VarsC = "C1")
res17$MedtImptM_all
res17$MedtImptM_list
res17$MedtImptM_table1
res17$MedtImptM_table2

12. Exit

FuncExit(PID = res$PID)

My31ren/exmedt