You can plot your relative gene expression levels wrt to haplotypes using standard non-parametric t tests, which will give you an assessment of the correlation between the two.

First, you should check the distribution of gene expression analysis. Is it normal or not? If normal, you can perform a general linear model to find the SNPs associated with expression level. Otherwise you should transform your data. Anyway, it depends on your hypothesis..are you looking for the single SNP influencing in the greatest way the gene expression, or are you looking for the best SNPs combination affecting gene expression? this is an important question. If you need a help, do not hesitate to contact me.

Hi Brijesh,

there is not a defined pipeline to do that. First of all are your qR-TPCR analyses done on what genes exactly...target genes or the SNP-containing gene itself?. Is your gene a transcription factor, a receptor, a cytoplasmic protein, a nuclear protein? These things need to be explained in your questions to narrow down the answers. Anyway, I'll try to address your question thinking about the expression of your own gene or targets genes.

The simplest option is, first check your SNPs have been described before (dbSNPs, http://www.ncbi.nlm.nih.gov/snp, or http://www.hgvs.org/dblist/ccent.html), if they have been described then look for some expression data associated with the SNP if available if not, then you were the first one and you can annotate that expression you found. In order to "correlate" you need more samples/cases to validate whether the expression of your gene of interest + SNP are always together. However, the actual expression of the gene is unlikely to be directly influenced by the SNP, it is the function and/or activity of the gene that will be affected (check this paper out: doi: 10.1007/978-1-60327-411-1_1.) Having said that, if your gene is also auto-regulating its own transcription then the SNP can also affect expression of its transcript...otherwise, quite rare.

If you are interested in seeing if your SNP-containing transcript affects the expression of other genes...you can do take a few target genes, measure their expression on various biological replicates and then use some statistic tests (Chi-sqaure will work ok as you have a nominal variable SNP-yes/SNP-no versus a quantitative variable, namely expression...BUT you can use something else) to correlate your SNP-status and Ct/expression values.

I hope this helps.

The simplest way to see a correlation is plotting your expression data in different genotypes or alleles and compare the measurements by ANOVA (genotype) and t-test (allele). You should test your variable for normality distribution, if not you must use non-parametric analysis.

There are some other multi-variable statistical analysis to correlating polymorphism with gene expression such as Stepwise Discrimination Function Analysis and Canonical Correlation. In some cases these two can help.

You may analyses the expression level of different genotype of a gene using t test. You also can use reporter gene assay to validate the polymorphism on the expression pattern.

Thanks for answers..I am checking single SNP of cytokines with its expression level.So do i need to analyse SNP of one cytokine of individual patient with its expression or it can be correlated between control vs patients group?

Your analysis depends on your hypothesis what you want to test. If you want to see how your SNP affects expression level you should plot expression levels versus genotypes or alleles. First check if the distribution of your data is normal or not, if not use transformation or non-parametric test. For normal distributed data or transformed you could use ANOVA with expression level as dependent variable and your genotype as independent one. You could also use linear regression models. Look at this paper it might help.http://www.ncbi.nlm.nih.gov/pubmed/20888525.

For case/control issue I would test overall expression versus genotype first to see if your SNP associates with gene expression. Then I would look if the distribution of genotypes and alleles is different between cases (patients) and controls. I don't think it's likely that SNP vs. expression would differ between patients and controls, but genotype/allele frequencies might.

We do a lot of this type of analysis. The simpliest thing to do is to stratify your samples by genotype (AA, AB, BB) and analyze expression of your various gene. If you are testing lots of SNPs you should correct for multiple testing.

We used mixes of genomic DNA (1:10, 2:8, 5:5...etc.) and found that while you can approximate levels, assessing levels is not very accurate. In my opinion, assessing levels is not really possible.

Genes have different expression level in different tissues. Thus, I think the tissue you selected for examing the expression and genotype is very important.

are you getting peptide expression or transcripts levels? Probably is the second, with normalization, and tissue specific, but there are so many factors influencing transcription that it is difficult to know. However, a simple association approach according to genotype would give an idea of the influence on the levels, but it is not possible to have definite conclusions, since you are not studying all variants...

Your question, as I think, seems only to see if genotypes correlate with certain genes expression. What you need to do just to make statistical analysis according to genotype If you have got the data . Firstly you should separate the data from patients and controls into several groups based on genotypes, for example, if only one SNP you detected, there are two alleles, A &B, and three genotypes AA, AB, BB, you should separate your data into 3 groups(genotypes). If you have two or more SNPs to detect, there will be more haplotypes(alleles) and genotypes. Then to compare if there are the significant changes of gene expression between the groups.

for expression analysis i have isolated m-RNA and prepared c-DNA.and i am checking only one SNP per gene...

The important issue to be taken care here is Sampling. If you have categorized the samples as Cases and Controls correctly and have genotyped for SNPs first you would have detected the variants.

To the same Cases and Control group if you have collected your PBMCs it is possible to isolate mRNA and perform qPCR experiments. I think you must have done that. Then you can conclude based on Genotype the corresponding gene expression for each sample...

please read e-QTL... I think you need only that.

Hi have been reading about eQTL, there are papers which mention that they do SNP and expression association test.

However, i failed to figure out how is it done.

What should be the format of the input files. I have a set of SNPs for each genes which I detect to be having differential allelic expression in hybrid .

I want to test if I can associates the SNPs to the expression variation to find potential eQTLs.

Brijesh were you able to do the association. Could you share your methodology !

Regards,

Ganga Jeena

http://www.ncbi.nlm.nih.gov/pubmed/20888525. as already mentioned above this link will help you for your work. Inspite of this if you need any help you can contact me anytime.

thanks.

Could you provide me your contact details !

Perhaps the presentation I gave a year ago to undergraduate students is any help:

http://www.dannyarends.nl/pdf/2011QTLintroduction.pdf

Hi,

So kind of you to share your presentation.

However the association test of multiple SNPs with allele expression is trick for me.

I tried Anova with following formula:

I provide the following data (1 column : expression counts; next columns, SNP_genoptypes:A=1, C=2, G=3 and T=4)

Ph SNP1 SNP2 SNP3 SNP4 SNP5

An1 300 1 4 4 2 2

Ler 50 3 1 3 2 3

An1XLer_An1 280 1 4 4 2 2

An1XLer_Ler 45 3 1 3 2 3

LerXAn1_An1 300 1 4 4 2 2

LerXAn1_Ler 40 3 1 3 2 3

and the following formula:

aov (Ph~SNP1+SNP2+SNP3+SNP4+SNP5,data=expression_and_SNPs)

However, it can not estimate for more than one SNP!

Terms:

SNP1 Residuals

Sum of Squares 92504.17 316.67

Deg. of Freedom 1 4

Residual standard error: 8.897565

4 out of 6 effects not estimable

Estimated effects may be unbalanced !

You can use the anova function for that,

the aov function in R is optimized for multiple phenotypes (and not genotypes)

linmodel

Hi,

thanks Danny!

I want to test for the most effective NSP or combination of SNPs with effect on expression !

However : even with anova I get only information for only 1 SNP, and no warning or message!

> linmodel (anova(linmodel))

Analysis of Variance Table

Response: Ph

Df Sum Sq Mean Sq F value Pr(>F)

SNP1 1 92504 92504 1168.5 4.37e-06 ***

Residuals 4 317 79

---

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Hmm, I should have a look at the data structure you use, this is an basic example:

snp1

x is the following data:

first column = expression count (Phenotype, Ph)

second column = genotype at first SNP for An1 and Ler

A=1, C=2, G=3, T=4

and next columns are the genotypes at further SNP locations

Ph SNP1 SNP2 SNP3 SNP4 SNP5

An1 300 1 4 4 2 2

Ler 50 3 1 3 2 3

An1XLer_An1 280 1 4 4 2 2

An1XLer_Ler 45 3 1 3 2 3

LerXAn1_An1 300 1 4 4 2 2

LerXAn1_Ler 40 3 1 3 2 3

If I specify the SNPs in alphabets, it doesnt except the formula with multiple SNPs itself!

I do not want the numbers at SNPs to be used for calculation of variance, I am trying to understand how the association between the genotypes and the phenotype can be established !

linmodel F)

SNP1 1 92504 92504 1168.5 4.37e-06 ***

Residuals 4 317 79

---

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

What template do you use for RFLP? I suppose that you use genomic DNA. With this material you cannot study gene expression, apart from prokaryotic organism. For gene expression, you need mRNA from the specific orgna, like pancreas for diabetic study.

Hi,

thanks,

I use RNA-Seq data to determine expression level per gene, per allele and SNPs!

Ok. so you study SNPs only for coding regions. That's ok.

segregate your genotypes and the corresponding expression data. Suppose the studied polymorphism is G/A so u have GG, GA, AA. Club all the expression data according to genotypes. Then use ANOVA with bonferroni multiple comparison. You can use GraphPad prism Software (much simple to use). You should get a meaningful correlation. Hope it helps........

Thanks Abhimanyu. It was very helpful.