dependecy

DESCRIPTION

@@ -5,7 +5,7 @@ Title: Independent component analysis for high-resolution mass-spectrometry base
 Author: Youzhong Liu <[email protected]>
 Maintainer: Youzhong Liu <[email protected]>
 Depends: R (>= 3.4.1)
-Imports: e1071, propagate, mixOmics
+Imports: propagate, mixOmics
 Suggests: ade4
 Description: ICA is an important alternative to classical statistical approaches for non-targeted metabolomics data. It extends the concept of regular correlation (e.g. in PCA, ASCA and PLS-DA) to statistical dependance by capturing higher order dependencies. However, its algorithm instability (output variations between different algorithm runs) and the biological validity of components have been overlooked when applied to complex metabolomics data. MetICA adresses these problems by gathering ICs estimated from multiple algorithm runs and from bootstrapped datasets, clustering them so as to find the most representative components. While evaluating the algorithmic stability, MetICA also suggests multiple criteria to select the correct number of components and to rank the extracted components.
 License: GPL3

NAMESPACE

@@ -3,7 +3,6 @@
 export(MetICA)
 export(MetICA_extract_model)
 export(validationPlot)
-importFrom(e1071,kurtosis)
 importFrom(grDevices,rgb)
 importFrom(graphics,barplot)
 importFrom(graphics,box)
@@ -13,6 +12,7 @@ importFrom(graphics,polygon)
 importFrom(graphics,title)
 importFrom(mixOmics,ipca)
 importFrom(propagate,bigcor)
+importFrom(propagate,kurtosis)
 importFrom(stats,as.dist)
 importFrom(stats,cor)
 importFrom(stats,cutree)

R/MetICA.R

@@ -47,7 +47,7 @@
 #' M1=MetICA(bacteria_peptides$X,pcs = 20,max_iter = 100,boot.prop = 0.3,max.cluster = 40,trends = T)
 #' # Generate validation plots along with geometric index calculation to help decide number of clusters
 #' validationPlot(M1)
-#' # According to the validation, we choose 10 components
+#' # According to the validation, we now choose 10 components:
 #' M2=MetICA_extract_model(M1,10,tops=7)
 #'
 #' @export
@@ -56,7 +56,7 @@
 #' @importFrom stats rnorm cor as.dist hclust cutree
 #' @importFrom mixOmics ipca
 #' @importFrom propagate bigcor
-#' @importFrom e1071 kurtosis
+#' @importFrom propagate kurtosis
 
 MetICA<-function(X, pcs = 15, max_iter = 400, boot.prop = 0.3, max.cluster = 20, trends = T, verbose=T){
 
@@ -281,7 +281,7 @@ MetICA<-function(X, pcs = 15, max_iter = 400, boot.prop = 0.3, max.cluster = 20,
     boot_eval_summary[[nb_cluster]]=dis_boot_no_boot
     defl_prop_summary[[nb_cluster]]=defl_prop
 
-    if (!all(total_number>30)){ # At least 30 estimates in each cluster
+    if (all(total_number)<=30){ # If < 30 estimates in all clusters
       max.cluster=nb_cluster
       message("Too few estimates in new cluster(s): max.cluster reset to: ",nb_cluster)}
 

README.md

@@ -3,6 +3,17 @@
 ## Context
 ICA is an important alternative to classical statistical approaches for non-targeted metabolomics data. It extends the concept of regular correlation (e.g. in PCA, ASCA and PLS-DA) to statistical dependance by capturing higher order dependencies. However, its algorithm instability (output variations between different algorithm runs) and the biological validity of components have been overlooked when applied to complex metabolomics data. MetICA adresses these problems by gathering ICs estimated from multiple algorithm runs and from bootstrapped datasets, clustering them so as to find the most representative components. While evaluating the algorithmic stability, MetICA also suggests multiple criteria to select the correct number of components and to rank the extracted components.
 
+## Install devtools (only if it has not been installed)
+
+Make sure you have a working development environment:
+* Windows: Install Rtools.
+* Mac: Install Xcode from the Mac App Store.
+* Linux: Install a compiler and various development libraries (details vary across different flavors of Linux).
+
+```R
+install.packages("devtools")
+```
+
 ## Installation from Github using R (with devtools)
 
 ```R
@@ -36,7 +47,7 @@ X = t(X) # Transpose the data since MetICA accepts samples x variables matrices
 
 ```{r}
 # Begin a MetICA simulation with 2000 estimated components in total. The samples are not time-dependent, so trend = FALSE. Numbers of clusters are evaluated between 2 and 10:
-M1=MetICA(X,pcs = 10,max_iter = 200,boot.prop = 0.3,max.cluster = 10,trends = F)
+M1=MetICA(X,pcs = 10,max_iter = 400,boot.prop = 0.3,max.cluster = 10,trends = F)
 # Users can confirm the number of pcs used for denoising if they think enough variance is explained, they can modify the number of pcs at this moment as well:
 ```
 ![choose](inst/Launch_MetICA.JPG)