A novel coarse-grained model is developed to elucidate thermodynamics and kinetic mechanisms of DNA self-assembly. It accounts for sequence and solvent conditions to capture key experimental results such as sequence-dependent thermal property and salt-dependent persistence length of ssDNA and dsDNA. Moreover, constant-temperature simulations on two single strands of a homogeneous sequence show two main mechanisms of hybridization: a slow slithering mechanism and a one-order faster zippering mechanism. Furthermore, large-scale simulations at a high DNA strand concentration demonstrate that DNA self-assembly is a robust and enthalpically driven process in which the formation of double helices is deciphered to occur via multiple self-assembly pathways including the strand displacement mechanism. However, sequence plays an important role in shifting the majority of one pathway over the others and controlling size distribution of self-assembled aggregates. This study yields a complex picture on the role of sequence on programmable self-assembly and demonstrates a promising simulation tool that is suitable for studies in DNA nanotechnology.