Non-orthogonal multiple access (NOMA) and reconfigurable intelligent surface (RIS) are expected to be key technologies in future wireless communication systems. With the ability of modifying the wireless channel, RIS has a great potential to improve the channel capacity and spectrum efficiency. However, the modification of wireless channel may change the optimal user pairing strategy of NOMA, which will cause large interference and decoding failure at the receiver. The joint optimization of user pairing and RIS beamforming in NOMA systems has not been studied yet. To address this issue, this paper considers a RIS assisted NOMA downlink transmission with one base station (BS), where the edge users are paired with the central users to share the same resource blocks, and RISs are deployed to serve the edge users. Our goal is to maximize the sum rate of edge users by jointly optimizing the user pairing, the passive beamforming at the RISs, the active beamforming and power allocation at the BS, subject to the users’ minimum rate and the total transmitting power. Since the formulated problem is non-convex and difficult to be solved, we propose an alternating iterative algorithm to obtain the suboptimal solution. Simulation results show that the proposed scheme can increase the sum rate of edge users by 5% and 7% compared to the other two baseline methods.