Hi Hui,

I guess you have to do qPCR for specific targets. You can either design specific primers or try Taqman qPCR where you can buy ready made probes.

Good luck! 

for single or few gene/s go for qPCR

The above answers do not address his questions directly.

It seems that Hui Song is saying that RNA-seq data is already available and he wants to estimate gene expression in a particular tissue he is interested in. I figure he also wants to know how to detect the non-expression of genes in other tissues (is RPKM zero?). This is quite different from the usual "differentially expressed genes" because a gene may be differentially expressed but might still be expressed in all tissues.

I think there are a few points to note:

1. For RNA-seq data, one would start with alignment of the short reads to infer what genes are expressed.

2. There are many tools for RNA-seq normalization. You perhaps want to normalize the data first before deciding on a suitable cutoff, the values should be high in the tissue you are interested in and near zero for the other tissue type.

3. Depending on how the alignment was done, I would also check if the value is zero, whether the reads were assigned to some close isoforms instead. Because the gene might not be tissue specific after all.

4. I also guess you need replicates to strengthen your analysis.

It's a tricky one because genes can be expressed at high levels in one tissue and at lower levels in other tissues (I think of this as tissue enriched) so setting a cut off of zero will only tell you about genes that are limited to one tissue but not enriched. You do need to normalise your data, set appropriate detection thresholds and decide how you define 'tissue specificity'. 

Another indirect way to assess tissue specificity of genes when working with tissues to use reference data from your organism where different tissues have been dissected and screened either with microarrays or RNAseq. This requires a model organism such as humans or flies. You can work out tissue specificity using a simple index tau, see papers below.

http://bib.oxfordjournals.org/content/early/2016/02/17/bib.bbw008.full

I work on flies and I have just done an RNAseq experiment on whole flies, but I calculated tau for my expressed genes using FlyAtlas reference expression data, and I was able to partition my genes into groups based on reference data among all the major fly tissues. There are many databases for different organisms.

Good luck!

Before any experiment in lab, I would check online databases beforehand, where gene expression is measured in different tissues. You might check out Gene Expression Atlas, GTEx, TiGER or BioGPS, to mention a few, and compare the expression values of your genes, after normalization, with those reported by these databases. This strategy would allow you to infer **rougly** the minimum expression levels of expressed or not-expressed genes in any tissue, for known genes. You might infer the localization of unknown genes, accordingly.

https://www.ebi.ac.uk/gxa/home

www.gtexportal.org/

bioinfo.wilmer.jhu.edu/tiger/

http://biogps.org

Hi Hui

With regards to your RNA-seq data I would suggest that you:

1) Make sure your data is also normalized for inter-library differences (aka not just seqeuncing deapth and gene length). You can use tools such as edgeR, Cufflinks or similar (https://bioconductor.org/packages/release/bioc/html/edgeR.html and http://cole-trapnell-lab.github.io/projects/cufflinks/). Then you can select a measure for defining what you will call tissue specificity - there is no de-facto way of doing this. Possible measures are cutoff on FPKM/RPKM, percent of total expression across tissues and if you want a strict set you also require that the gene is significantly differentially expressed between your tissues (using the same tool you used for the inter-library normalization.

2) You compare to other types of data - for your case the easiest is probably to use slideBase which is designed especially for this purpose and allows you to analyze data from all the major sources (GTEx, Human protein atlas, FANTOM etc). You can find it at http://slidebase.binf.ku.dk/

Hope this helps

/Kristoffer