Accurately determining the masses of supermassive black holes is crucial to understand their evolution and the establishment of their relationship with their host galaxy properties. Beyond the local universe, the single epoch mass estimation method provides a simple procedure to estimate black hole masses in large spectroscopic samples of type-1 active galactic nuclei. The method assumes virialized motion of gas in the close vicinity to the active black holes, traced through broad emission lines. However, because of the assumption of a universal virial factor, this procedure has uncertainties associated with the unknown distribution of the gas clouds. Here, using a sample of 39 quasars observed with the VLT/X-shooter spectrograph, we compare alternative estimations of black hole masses determined from the properties of the accretion disk emission around the black hole with the single epoch virial mass estimations. We find that the virial factor is inversely proportional to the observed width of the broad emission lines. This result implies that current virial masses can be miss-estimated by up to a factor of 6. Our analysis indicates that either the effect of line-of-sight inclination in a planar distribution of the broad line emitting gas or the radiation pressure perturbations to the distribution of gas can reproduce our findings.