no method for coercing this S4 class to a vector for utilization of mclust

Question

I'm trying to use the mclust method on an .FCS format file (which is a flow cytometry format file) and I read this file into R as flowFrame object.

install.packages("openCyto") # since the old version sefaulted my R session
library( openCyto )
library(  flowCore)
library(  mclust)

 trial1=read.FCS("export_Alcina TregMAIT_AV 10-1974 P1_CD4.fcs") 
 a=as.matrix(trial1)

Editors note: some of these are Bioconductor packages and you should install according to the help pages for that environment.

However, mclust does not accept the .fcs file as a matrix & I tried to convert it to a matrix with the function as.matrix, and I get this error:

Error in as.vector(data) :
  no method for coercing this S4 class to a vector

I've found similar questions where they explain you have to add importMethodsFrom(S4Vectors,as.matrix) into the NAMESPACE of mclust, which I did. I also did importMethodsFrom(BiocGenerics,as.vector) in the NAMESPACE of mclust. However, I'm still not able to use mclust.

P.S. any advice or reading would be appreciated!

If, anyone knows other clustering methods that use GMM model that could accept .FCS format without converting, I'd be very happy.

Answer 1

I've edited your question to show what you should have done originally and also didn't do later (instead of including code in a comment you should have responded by then editing the question as was specifically suggested.) My response is based on the first example in flowCore::read.FCS (since you also did not include a pointer to the dataset you were loading from disk) so rather than "trial1" I will be referring to the "samp" object I get running that code.

The "samp" object is now returns this from class and str:

> class(samp)
[1] "flowFrame"
attr(,"package")
[1] "flowCore"
str(samp)
Formal class 'flowFrame' [package "flowCore"] with 3 slots
  ..@ exprs      : num [1:10000, 1:8] 382 628 1023 373 1023 ...
  .. ..- attr(*, "dimnames")=List of 2
  .. .. ..$ : NULL
  .. .. ..$ : Named chr [1:8] "FSC-H" "SSC-H" "FL1-H" "FL2-H" ...
  .. .. .. ..- attr(*, "names")= chr [1:8] "$P1N" "$P2N" "$P3N" "$P4N" ...
  .. ..- attr(*, "ranges")= num [1:8] 1023 1023 10000 10000 10000 ...
  ..@ parameters :Formal class 'AnnotatedDataFrame' [package "Biobase"] with 4 slots
  .. .. ..@ varMetadata      :'data.frame': 5 obs. of  1 variable:
  .. .. .. ..$ labelDescription: chr [1:5] "Name of Parameter" "Description of Parameter" "Range of Parameter" "Minimum Parameter Value after Transforamtion" ...
  .. .. ..@ data             :'data.frame': 8 obs. of  5 variables:
  .. .. .. ..$ name    :Class 'AsIs'  Named chr [1:8] "FSC-H" "SSC-H" "FL1-H" "FL2-H" ...
  .. .. .. .. .. ..- attr(*, "names")= chr [1:8] "$P1N" "$P2N" "$P3N" "$P4N" ...
  .. .. .. ..$ desc    :Class 'AsIs'  Named chr [1:8] "FSC-H" "SSC-H" NA NA ...
  .. .. .. .. .. ..- attr(*, "names")= chr [1:8] "$P1S" "$P2S" "$P3S" "$P4S" ...
  .. .. .. ..$ range   : num [1:8] 1024 1024 1024 1024 1024 ...
  .. .. .. ..$ minRange: num [1:8] 0 0 1 1 1 0 1 0
  .. .. .. ..$ maxRange: num [1:8] 1023 1023 10000 10000 10000 ...
  .. .. ..@ dimLabels        : chr [1:2] "rowNames" "columnNames"
  .. .. ..@ .__classVersion__:Formal class 'Versions' [package "Biobase"] with 1 slot
  .. .. .. .. ..@ .Data:List of 1
  .. .. .. .. .. ..$ : int [1:3] 1 1 0
  ..@ description:List of 164
  .. ..$ FCSversion                : chr "2"
  .. ..$ $BYTEORD                  : chr "4,3,2,1"
  .. ..$ $DATATYPE                 : chr "F"
  #----- output truncated -----------

So "samp" is not a rectangular objects in any sense but rather a complex list with lots of the associated information in attributes. My guess is that you want the information in the @ exprs node which is a matrix.

A further difficulty is that there is no function mamed mclust in the mclust package, although looking at ?mclust we do see an example demonstrating the use of an Mclust function. R is unforgiving in its insistence on correct capitalization of function names.

Mclust(exprs(samp)[1:100,])
#-----------
'Mclust' model object:
 best model: ellipsoidal, equal orientation (VVE) with 4 components