Tissue specific enrichment

[pyrot...@gmail.com]

Hi Kyoko and team,

What a great job and tool! Thanks so much for developing this.

I have a question about the tissue specific enrichment in the SNPtoGene function. I read the manuscript, and in there, there was a tissue enrichment for Genetofunction. Is it a similar process? Am I right in thinking that this is the general pipeline for the tissue specific enrichment in the SNPtoGene function?

1. T-test gene expression in focal tissue against the average gene expression in all the other tissues

2. Significant genes are genes with fold change > 0.58 and Bonferroni P < 0.05

3. Enrichment is a standard MAGMA gene set analysis

Could you also please let me know why you chose a fold change > 0.58 for identifying significant genes?

Many thanks

Varun

[Kyoko Watanabe]

Hi Varun,

Thank you for using FUMA, I think you might be bit confused about tissue enrichment tests between SNP2GENE and GENE2FUNC.

In the SNP2GENE, MAGMA tests tissue specificity by using gene expression value as a covariates conditioning on average expression across tissue types (so this is different from gene-set analyses). In other words, we test positive correlation between significance of associations and expression. We do not define "gene-set" here, instead we use global distribution of associations and gene expression.

In the GENE2FUNA, as you mentioned, I performed differential gene expression analyses (t-test, a certain tissue against all other) and create DEG gene sets. Then perform hypergeometric test to see if the list of genes of interests are enriched in DEG.

About fold change, the threshold is log2 scale, so 0.58 represent actual fold change 1.5. In FUMA, we use three different DEG, up-DEG (log2FC > 0.58), down-DEG (log2FC < -0.58) and DEG (|log2FC| > 0.58).

Hope everything is clear,

Best,

Kyoko

[pyrot...@gmail.com]

Hi Kyoko,

Ah, I see! So, if I want to replicate this on MAGMA, would it be:

magma --gene-results [GENE_RESULTS].genes.raw --gene-covar ["Insert covariate file with gene ID in the first column and tissue specific expression in the second column"] --out [OUTPUT_PREFIX]?

Thanks a lot,

Varun

[Kyoko Watanabe]

Hi Varun,

Yes, you are almost correct, but for gene covariates,

--gene-covar <file with geneID in first column and log2(RPKM+1) for all tissue types + one more column for average expression> onesided=greater condition=<column name of average>

Best,

Kyoko

[pyrot...@gmail.com]

Ah fantastic! Thanks Kyoko!