Recommender systems (RS) aim to predict items that users would appreciate, over a list of items. In evaluation of recommender systems, two issues can be defined: accuracy of prediction which implies the satisfaction of the user, coverage which implies the percentage of satisfied users. Collaborative filtering (CF) is the master approach in this domain, but still has some weaknesses especially about coverage. Trust-aware approach is today another promising approach in RS within social environments, whose prediction exceeds the quality of (CF). In this paper we propose several strategies to hybridize those both approaches in order to improve prediction quality, in the term of accuracy and coverage.