Normalization - Tissue specificity analysis

Question

[Filago]

Hi guys, I want to perform tissue specificity analysis (z-score etc.) on publicly available datasets like GTEx. So for each gene I want to calculate how tissue specific it is.

However, first I need a normalization strategy for the raw counts. A lot of papers/tools in this field simply work with the median tpms per gene per tissue, however I think this not a proper solution and that a between sample normalization should work better.

My plan is to generate a matrix, which contains all samples, from all tissues (columns) and all genes (rows) and to perform a TMM-normalization on it (not explicitly considering tissues as if all would come from the same tissue). Does anyone of you have experience in that field and thinks, that this strategy is a proper one?

Thanks, Andreas

Answer 1

[Istvan Albert]

A little bird told me of this package:

https://github.com/ATpoint/CreateGeneSignatures

This package implements a simple yet effective rank-based strategy to produce celltype-specific gene signatures based on RNA-seq differential expression results.