Antimicrobial peptides (AMPs) interact with bacterial membranes to induce membrane disruption and protect host cells from bacteria through antimicrobial action. Although the mechanisms of the antimicrobial actions have been studied in experiments and simulations, understanding of the mechanism at the molecular level has not been developed due to the complexity of the antimicrobial action. In this study, we conducted a series of molecular dynamics simulations using all-atom (AA) or coarse-grained molecular dynamics (CG-MD) simulations to reveal actions of melittin, a typical AMP derived from honey bee venom, on lipid membranes. We performed AA-MDs of lipid membranes in the presence of multiple melittin peptides to estimate the free energy required to form a membrane pore. The AA-MD results indicated that membrane pore formation processes, namely antimicrobial actions, induced by melittin were likely to occurred in a concerted way involving multiple melittin peptides. To investigate the effect of protein-to-lipid (P/L) ratios on the antimicrobial action, we also carried out CG-MDs of melittin peptides adsorbed on a lipid membrane at different P/L ratios. To this end, we developed pSPICA, a chemically accurate, quantitative CG model designed based on polar CG water molecules. The CG-MD simulations using the CG model showed that three different modes were observed depending on the P/L ratio and the local arrangement of peptides on the membranes: toroidal pore formation, lipid extraction and budding, and bursting. Our observations are consistent with both our AA-MD results and experimental findings. In addition, we revealed that the P/L ratio and local arrangement of melittins were important factors in the initiation of and competition among the pore formation mechanisms.