Clinically relevant sensitivity in the nanogram/ml range can be reached using LC-SRM analytical chain. Meanwhile, CPTAC study has emphasized estimation variabilities, making major concern on the ability to assess biological variability. Indeed, it is not desirable that instrumental variability hides biological one. Protocol standardization is expected to improve this but will not solve variability after sample injection in the analytical chain. In fact, to get a robust estimation, several transitions are acquired, each one containing protein concentration information but also transition specific variability such as unexpected contaminants, or peptide fragmentation effects. Typically, average or median value of the protein concentration estimated from each transition is computed. Outliers' removal or ad-hoc rules to suppress low quality signals are also used. Such techniques combine information obtained on several acquisitions, thus averaging instrumental noise effects. Variability effects cannot be accessed independently and are lumped in estimated concentration distribution. We propose to introduce a statistical approach based on the distribution of both instrumental and biological parameters. The objective is to process independently each SRM acquisition without preprocessing or data removal. With the reduction of instrumental effects on the estimation of true protein concentration, a significant improvement of day-to-day, intra laboratory and inter laboratories variability is expected. We propose to introduce a relevant hierarchical model for the SRM chain within a Bayesian statistical framework.