Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 1 down vote favorite

I have expression values (log2) for 200 genes in two conditions treated and untreated and for each condition I have 20 replicates. The dataset looks like this:

Gene      UT1             UT2             T1              T2  
DDR1      8.111795978   7.7606511867    7.9362235824    7.5974674936
RFC2      10.2418824097 9.7752152714    10.0085488406   9.5723427524
HSPA6     6.5850239731  6.7916563534    6.6883401632    7.3659252344
PAX8      9.2965160827  9.2031177653    9.249816924     8.667772504
GUCA1A    5.4828021059  5.3797749957    5.4312885508    5.1297319374

I have shown only two replicates for each sample in the sample data.

How can I calculate similarity in gene expression for each gene in two conditions and rank them from highest to lowest (in terms of similarity)?

I am looking for a solution in R or python. The cor function in R does not give me what i want.

share|improve this question
1  
What do you mean when you say that you want to test for correlation? Do you mean that for each of the 200 genes you want to do a separate test that the correlation between the treated and untreated groups is different from 0? That would mean doing 200 tests and would have the usual problem of multiplicity (which is a serious problem when doing so many tests). On the other hand you might want to test for a single correaltion between groups ignoring what the specific gene is. – Michael Chernick Jun 15 '12 at 11:35
This would then involve only 1 test but would require the assumption that the gene doesn't matter(1.e. correlation is the same for each gene). – Michael Chernick Jun 15 '12 at 11:35
Hi Micheal, I have edited the question. I think t-test will do the trick ???????? – Angelo Jun 15 '12 at 11:45

2 Answers

up vote 1 down vote

For the modified question I assume Angelo that you will use the estimated correlations to rank the genes accoding to similarity. Then you want to test whether or not the rankings are similar for the two groups across the genes. To do this you could you the sample Pearson correlation to get the estimates for each of the genes and to rank them. Then use a non-parametric test of correlation (e.g. Kendall's tau) to test for similarity between group. The pairs for the test would be the rank in group 1 paired with the rank in group 2 for each specific gene. I don't think you would want to do a t-test on the rank differences. A nonparametric test such as the Wilcoxon rank sum test on the rank differences might make sense though.

share|improve this answer
Could you please suggest a stats tutorial for me (preferrably online)? – Angelo Jun 15 '12 at 12:12
up vote 1 down vote

Have you read some specific literature on gene expresion?. I think that a multivariate approach since your response variables are the expresion of 200 genes, and you have 20 replicates per condition (fixed factor). There are a lot of similary indices (see) available. If you are interested to test whether both groups differs in gene expresion, you can test this by means of a PERMANOVA (Multivariate permutation analysis). This package has plenty of different similary index to choose from (also Pearson correlation). If you detect statistcal signficant differences between those groups, you can examine which gene/genes are responsable for the difference between the two conditions. Excuse my english!

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.