Impute protein expression matrix

Question

[mropri]

Hi,

I have a matrix of protein expression that has some missing values. I know its MNAR (missing not at random). However, I am having trouble using the impute.QRILC command to impute the matrix as I get the same matrix as a result. Would someone be able to me understand why the matrix is not imputing? My code is the following:

y_qrilc <- impute.QRILC(norm_prot)[[1]]

norm_prot is my protein expression matrix

Any help is appreciated.

[Jeremy]

Are you getting any error messages? I had issues installing the imputeLCMD package in both Linux and Windows. Can you use another imputation method, such as k-nearest neighbor?

[mropri]

Yes, I was getting error messages. However, I was able to install imputeLCMD package and run impute.QRILC but was getting the same matric back. Would k-nearest neighbor be acceptable method for values that are not missing at random?

Answer 1

[Jeremy]

Multiple imputation is recommended for MNAR data (see the link to the paper below). You can use the mice package in r to do this.

Multiple Imputation Article