Relay-assisted cellular network is one of the most promising architectures for the next-generation mobile cellular system, which is envisaged to support high-rate multimedia services in a wide variety of environments: indoors, outdoors, low-mobility, high-mobility, etc. This work aims to investigate the theoretical performance of downlink transmissions of relayassisted cellular networks in the multi-cell environment with optimized system parameters. A genetic-algorithm based approach is proposed for joint multi-cell optimization of system parameters including locations of relay stations, path selection, reuse pattern and resource allocation to maximize the system spectral efficiency. Two types of quality of end-user experience (QoE) (fixed-bandwidth allocation and fixed-throughput allocation) are investigated along with two path selection schemes (spectral efficiency-based and SINR-based). Numerical results show that with the deployment of relay stations, the system performance is significantly improved over the conventional cellular networks.