I would like to compare the transcript profiles of some genes in several genotypes to find some relationship between the expression of these genes and the genotypes.
It depends on the number of repetitions for each category. If there are triplicates for example, the correlation based on Pearson coefficient is inappropriate, as the degree of freedom is n-2, i.e. 1. Euclidean distance is better. If there are 6 or more repetitions, Pearson's coefficient gets more reliable. In both cases, UPGMA give satisfactory clustering. Using Ward's method, you cannot choose Euclidean or Pearson. I personnaly use PAST of Hammer et al 2001, a free, intuitive and convenient program for statistics. Its pdf manual is useful. Another way to consider your data is to perform a PCA (principal component Analysis), easy to interpret, explained for exemple in :
M. Ermonval, D. Petit, A. Le Duc, O. Kellermann, P.F. Gallet,
Glycosylation-related genes are variably expressed depending on
the differentiation state of a bioaminergic neuronal cell line:
implication for the cellular prion protein, Glycoconjugate J. 26
Dear Daniel, thank you for your helpful answer! I learned something useful. I focused so much on cluster analysis that I forgot to notice PCA as another appropriate method to evaluate my results. This PAST software you mentioned is quite user-friendly. Thanks so much!
My answer is late for Mária but could help some others. I am agree with Daniel Pierre Petit regarding the use of Pearson r or Euclidean distance. However with both it is possibel to use Ward´s method. Wardd´s method is an amalgamation or linkage rule while Euclidean or Pearson r are distance mesures of cases and variables. The adventage for Ward´s versus other linkage methods is you are able to obtain very equilibrated clusters and clearly appreciated the signification of the groups