The understanding of pathophysiology of mental diseases is mandatory for the strengthening of precision psychiatry, since it can helps in prognosis, diagnosis, treatment, and monitoring of patients. In this context, the use of molecular biomarkers and sleep quality as tools seems to be promising. Among psychiatric illnesses, the mood disorders, especially major depression (MDD), have gained attention by the scientific community due to their increasing prevalence and morbidity. Initially, this study evaluated the sleep quality (Pittsburgh Sleep Quality Inventory: PSQI) and the levels of different peripheral molecules, such as: plasma C-reactive protein (CRP), the neurotrophic factor derived from the mature brain (mBDNF), serum cortisol (SC) and response to salivary cortisol (CAR), for compare them between a group of patients with major depressive disorder (n = 58) and a control group of healthy volunteers (n = 62). While patients in the first episode (de novo) (MD n = 30) had significantly higher CAR and SC levels than controls (n = 32), they showed similar mBDNF concentrations. Patients with treatment-resistant depression (TRD n = 28) had significantly lower levels of SC and CAR, and higher concentrations of mBDNF and PCR than controls (n = 30). Still, the severity of depressive symptoms and the poorest quality of sleep were negatively correlated with SC and CAR, and positively for mBDNF. Then, we evaluated the potential of these variables as biomarkers of MDD diagnosis and disease progression, for it we tested some multimodal mathematical models using multivariate logistic regressions and the RoC (Receiver Operating Characteristic Curve) curve. None of the tested diagnostic models had a higher accuracy than the Hamilton-6 depression rating scale (AUC = 0.99). The best model (AUC = 0.99) of disease progression included: PSQI, CAR, SC and mBDNF. In summary, these findings indicate that the type, intensity and direction of biological changes in MDD differ according to the progression of the disease. Moreover, the impact of these biological changes is mainly relevant for prediction of the progression of MDD, but not for its diagnosis. Thus, the present results help in the comprehension of pathophysiology of MDD and provide a foundation that can contribute to future studies that aimed at the development of mathematical models of MDD biomarkers that can be commercially available and used in the clinical practice of precision psychiatry.