Guys, I have microarray data and I ended up at a dead end. I have the DEGs in each treatment, and I know which pathways are affected in each treatment, but turns out it isn't enough due to the complexity of my analysis. The thing is, my dataset is composed of 5 treatments: 1. a positive control, 2. a toxin and 3, 4, 5, major metabolites x,y and z of this toxin, each one in 3 times points (6.12 and 18h of incubation with cells). I was thinking about using an approach to be like "the response of the toxin after 12 h is 23% similar to the response of the metabolite x after 6h".
I tried to use pearson correlation, but it only gives me the R value, which I can't turn into a percentage.
I also have the cosine similarity distance, but it's pretty similar to pearson. I also tried linear regression, binary correlations, but I ran out of ideas. Any suggestion, or paper or, mathematical advice I could use?
Thanks in advance.
Maybe this publication may provide some useful info.
Analyzing time series gene expression data
http://bioinformatics.oxfordjournals.org/content/20/16/2493.short
For time-series experiments I have used STEM, it is quite user-friendly and it has GO support, which works great at least for Arabidopsis thaliana: http://www.cs.cmu.edu/~jernst/stem/
As a suggestion for additional analyses, you could also try to construct gene co-expression networks, compare your treatments and identify significant differences among them. Potentially, that could give you some clues about the direct targets of your toxin.
Co-expression network analyses have been useful for identification of transcription factors regulating biosynthetic pathways in Arabidopsis: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1849962/
It has also been used for microarray experiments in humans: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2847743/
I have not used it, but this software looks quite interesting for this type of analysis:
http://www.ncbi.nlm.nih.gov/pubmed/21756334
Hi Gustavo, I'll try to use STEM. Concerning the co-expression network, I determined all the different pathways/networks affected by each treatment. Since I'm working with a toxin and its metabolites, what I expect to see is late responses of the toxin will be similar to the early responses of the metabolite X. Furthermore, each metabolite affect different cluster of genes, i.e. genes involved in oxidative stress, or apoptosis or AhR pathway. But yes, I'll take a look at these co-expression network analyses, maybe they'll give me an insight. Thanks!
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis