how to calculaute Tissue Specificity of genes

Question

[Bioiris]

salut,

I have a count table , and I need to calculate the tissue specificity of genes

table(A) ————————

gene_id sample1 sample2 sample3 . ……..sample20 g1 23 0 1 3 g2 35 1 4 8 g3 4 6 70 0 g4 98 12 7 6 g5 60 2 8 12 . . . g1907 13 0 0 0

is there a tool that calculates tissue specificity ??

Thanks.

[spvensko]

There are multiple methods to estimate tissue specificity. Here's an article that compares some approaches: http://bib.oxfordjournals.org/content/early/2016/02/17/bib.bbw008.full

[BioinfGuru]

I am doing the same thing. I read that paper last night. Very fast read ...recommends the TAU method. But also points to some pitfalls to avoid.

Other papers from this year:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4932457/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4968975/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4916594/

What tissue are you working on?

[Hranjeev]

Try Jaccard index, I have used them with RNAseq data before.

Answer 1

[Satyajeet Khare]

You can use Specificity matrix in cummeRbund. S=1 if the gene is expressed exclusively in that sample.

Answer 2

[Charles Plessy]

I like the richness index that is implemented in the "rarefy" function of the "vegan" package for R. For example, if one calculates the richness index on a scale (subsample size) of 10 for a given gene, it is basically asking the question of how many samples would seen be expessing that gene if the total number of counts for that gene were 10. The lower the richness, the higher the tissue specificity. To normalise the expression table before calculating the richness, I downsample all the libraries at the same level. Note as well that a meaningful richness index can not be calculated on a scale that is lower than the total number of counts for a given gene. Thus, in the example above, genes with less than 10 counts in total over all the sample should be excluded.