The G3BP2 locus has been discovered and studied in [1]. In a recent set of quantitative reverse transcription polymerase chain reaction (qRT-PCR) experiments, we have been trying to study the G3BP2 transcripts and their relationships with their genomic neighborhood. The gene expression experiments we set up involved normalizing housekeeping genes, several different transcripts, several conditions of interest (different cell lines, tumor versus normal tissue), technical replicates, missing data and small sample sizes in all resulting cells. The scenario is suitable for the analysis via Bayesian hierarchical models using as response variable the continuous interpolation of cycles in the qRT-PCR, which is naturally taken to be lognormal. The construction of few normal hierarchical models will be discussed, critical points will be illustrated and a comparison with standard methods will be made. Currently, the state of the art in analysing qRT-PCR data is based on ad hoc user-friendly software (such as REST) or on linear mixed models. We will show how the alternative analysis based on statistical Bayesian networks can be flexible enough to address the important issue of estimating gene-gene interactions.