I would like to get some hints for the following problem, I am using R language, and trying to find a statistical model for my biomedical question. The question is as follows:

I have two matrices:

B: 990*255 mixed matrix, AUC values (cell-line* drug)- between 0 and 1-some values are missing

A: 78*255 mixing matrix (drug * target)- {only 0 and 1, 1 is when one drug is targeting to one protein(target), and 0 otherwise }

Now I would like to use deconvolution to derive the connection between cell-lines and targets, namely, for a cell-line and a target we would like to find a value which shows whether that target is critical for that cell line. It means finding key targets for each cell line.

X: 990*78 source matrix (cell-line target)-unknown

Problem of X*A=B.

I have tried to use blind source separation methods, but I need also to use the mixing matrix A which contains important information.

My first question is how to deal with missing values without using imputation.

Second question is how to approach to this problem? Which deconvolution method would be suitable for this problem?


It might be possible to use semi blind separation method. The idea is that the matrix A which is the mixing matrix is only partly known so in principle we want to estimate that too. But we need to take the primary binary A matrix into account as well and then try to use semi blind separation method.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.