Can singscore handle missing values

[evrama]

I am exploring your tool to score a bidirectional signature in my proteomics data. Can singscore handle missing values in the expression matrix? I tried and I am getting NAs in the score output

[bhuvad]

Hi @evrama,

Singscore cannot be used with missing values in the matrix as it is as it was primarily designed for use with transcriptomic data. For proteomic data, you could use a workaround whereby you score each sample independently while removing NAs. The method would hold mathematically if you do so. I am keen to integrate this in the future too since proteomics data is becoming more readily available now. I will work on adding this as a feature ASAP (though it may take a few weeks to get it done). I would recommend using the workaround I suggested in the meantime. I will notify on this post once that feature is integrated.

Cheers,
Dharmesh.

[AJ714]

Hi,
and thanks for a great tool! I was wondering if there have been any advances with respect to the handling of proteomic data? As by now, I am following the recommended workaround mentioned above, removing NAs for each sample before ranking and scoring. However, having a rather large amount of samples, I would really appreciate to find a more integrated solution in singscore for this kind of data.
Cheers,

[pcantalupo]

Has this been fixed in singscore yet? I have alot of NA in my data. Here are my samples and number of missing data:

                  Genes_NA Genes_OK percOK
4dpi_Mock             7461    40976   84.6
6dpi_Mock             7461    40976   84.6
Fed4dpi_1dpf_Mock     7461    40976   84.6
Fed4dpi_2dpf_Mock     7461    40976   84.6
Fed4dpi_4dpf_Mock     7461    40976   84.6
Fed6dpi_1dpf_Mock     7461    40976   84.6
Fed6dpi_2dpf_Mock     7461    40976   84.6
Fed6dpi_4dpf_Mock     7461    40976   84.6
mock_run6            34896    13541   28.0
PTref                21132    27305   56.4
aPT                  21132    27305   56.4
cycEPI               21132    27305   56.4
dPT                  21132    27305   56.4

The worst sample is 'mock_run6' with 72% missing expression data. The reason is that I'm merging 3 RNAseq and scRNAseq experiments that can have different sets of gene symbols as well as some samples with low depth.

So, do you recommend that I run singscore individually on each sample after removing NA values? How comparable are the singscores across samples if I do this?

Thank you