Biology Stack Exchange is a question and answer site for biology researchers, academics, and students. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I was reading a paper related to bioinformatics where it uses the drug response on the cancer cells and the gene expression of the individual cells are studied to find any useful insights. Specially, using the gene expression of the cells a predictor of the drug response is created.

They have stated that just using the correlation between the gene expression and the drug response might not be a good predictor. But the genes interact through signaling pathways to drive a particular drug response.

What these guys have done is like used PCA on the gene expressions of the cancer cells to use the components which preserve the greatest variance.

Actually, I didn't get what they mean by probe-by-background interaction and how it is calculated.

Can anyone please explain. I googled for a while but didn't get it.

Here are some quotes from the paper where the term is used

Towards this end, we have compared how well drug response can be predicted by simple statistical models, which either directly relate probe and background networks to drug response or consider probe-by-background network interactions.

To generalize this approach, the term ‘probe’ could be replaced by individual transcript expression levels measured through other gene expression methods. Similarly, ‘background networks’ and principal components are used interchangeably. Generally, ‘background networks’ could be represented by any data reduction method that summarizes the expression of a gene network. We demonstrate that probe-by-background network interactions significantly enhance drug response predictions, over and above the predictive power garnered through utilizing individual probes and background networks alone.

share|improve this question
cross-posted with BioStar: – Michael Kuhn Apr 8 '13 at 7:29
Posting the paper here is likely to get everyone into trouble, as the content is owned by Nature Publishing Group. I added a link to your post - at least a few of us do have institutional access. Also, I do not expect anyone to try to read the article from screenshots. – dd3 Apr 8 '13 at 21:43
Rather than including screenshots, can you provide a few examples of where the authors use the term that you are unsure about? Just a sentence or two, quoted, would be fine and within fair use, I think. – kmm Apr 9 '13 at 0:35
I rolled back your edit. We do not want screenshots, as that is very likely a copyright violation. My suggestions was to quote several instances where the term is used. – kmm Apr 9 '13 at 2:38

The "probe by background" interaction is the response of different probes as a function of background gene expression. For example, depending on which of the 6 backgrounds a probe is in, the drug response may go up or down. Probes as a function of background is probably easier to imagine than background as a function of probe (which is equally valid). For 39,115 probes and 6 background networks, there are 234,690 interactions.

In a technical sense, "probe by background" is the interaction term in the linear model that the authors fit for each drug. The details of the analysis are in the supplemental methods. The model they fit is

Drug_Response ~ Probe + Probe * Background + Background

which translates to "Drug response is modeled by Probe, the interaction of Probe and Background, and Background". The middle term (with the *) is the interaction term. Here is a page that helps to explain how to understand significant interaction terms in linear models.

share|improve this answer
I didn't understand how it is calculated. I mean probes are the original probes. Background are formed by applying PCA and taking the best 6. I didn't understand how this interaction term is calculated. How do I know which background a probe is in – user34790 Apr 9 '13 at 13:40
I didn't understand how it is calculated. I mean probes are the original probes. Background are formed by applying PCA and taking the best 6. I didn't understand how this interaction term is calculated. How do I know which background a probe is in. Also I didn't get why it is called background. I mean they have just applied PCA to capture the decorrelated basis which capture the most variance. Why is it called background? – user34790 Apr 9 '13 at 13:47
I am wondering if they used the ttest values for the interaction term? – user34790 Apr 9 '13 at 14:38
I just found out that it is just the product of the probe values and the background values. However, they have mentioned probe-by-background term(paired t-test P-value=7,5*10^-5, paired Mann-Whitney test P-value=5.2*10^-5). What is meant by this? – user34790 Apr 9 '13 at 16:56

Your Answer


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.