Predicting phenotype based on the expression of genes

Question

[zizigolu]

Hi,

Supposing we have measured cholesterol level in control and treated mice (treated by a drug). We also conducted microarray between control and treated mice. Is it possible to predict the responses shown in cholesterol level based on the expression of genes???

By googling nothing came in my mind

Answer 1

[Kevin Blighe]

Sorry, from our other old thread where i just picked up your added comment: C: Gaussian modeling of GWAS data

Yes, that is possible. In that case, you would use a linear model with cholesterol as the y variable:

lm(cholesterol ~ gene1)
lm(cholesterol ~ gene2)
lm(cholesterol ~ gene3)
...
lm(cholesterol ~ geneX)

Then take the statistically significant genes and put them together into a preliminary 'final' model that would require yet further testing:

lm(cholesterol ~ gene4 + gene10 + gene678)

( for further testing of the preliminary final model, see here: A: Resources for gene signature creation )

lm() will assume Gaussian.

-------------------------------

You could also just use glm() and specify Gaussian as the family, which would be the same (but can differ if you specify a particular link function with the family).

glm(cholesterol ~ gene1, family=gaussian)

[zizigolu]

Thanks a lot, always helpful