Depression is a very common disease. Approximately 10% of people in the Western world experience severe depression during their lifetime and many more experience a mild form of depression. It is commonly believed that depression is caused by malfunctions in the biological system constituted by the hypothalamus-pituitary-adrenal (HPA) axis. We pose a novel model capable of showing both circardian as well as ultradian oscillations of hormone concentrations. We show that these patterns imitate those observed in the corresponding data. We demonstrate that patient-specific modelling shows its ability to make diagnoses more precise and to offer individual treatment plans and drug design. Efficient and reliable methods for parameter estimation are crucial. Presently we are investigating how well the Metropolis-Hastings Algorithm of the Bayesian Markov Chain Monte Carlo (MCMC) method for estimating the parameters is working and we are about to do the same using iteratively refined principal differential analysis (iPDA) or the approximated maximum likelihood estimate (AMLE). Preliminary results for both methods are promising. The next step is to investigate which parameters there are responsible for which pathologies by statistical hypothesis testing.