Disulfide modified lipopeptide assemblies with an arginine-rich dendritic periphery provide a promising platform for effective gene transfer. Dendritic arginine peptides that mimic the cell-penetrating peptides of a virus envelope are vital for complexation, interaction with physical barriers, and final gene release. Here, we report three lipopeptides with different-generation dendritic peripheries (R₁LS, R₂LS and R₃LS), each of which contains a dioleoyl-L-lysinate hydrophobic tail. Such molecules were proven to self-assemble in aqueous solution with different morphologies, sizes, and surface zeta potentials. R₂LS and R₃LS assemblies showed spherical and spindle shapes with zeta potentials of 27.2 and 32.8 mV, respectively. They exhibited complete condensation of pDNA at a low N/P ratio, while R₁LS assemblies displayed a fiber pattern with a relatively low electric potential of 10.9 mV with poor DNA binding ability. In a cellular viability experiment, R₁LS and R₂LS have no significant cytotoxicity even at high dosage, while R₃LS showed conspicuous toxicity. As a gene vector, R₂LS presented high gene transfection efficiency either in the presence or the absence of serum, which was 58.7% greater than liposome 2000 and PEI in the condition of 10% fetal bovine serum for HeLa cells. While R₃LS showed good results just without serum and R₁LS was unserviceable in all situations. Moreover, molecular dynamic simulation was exploited to analyze the kinestate of the signal molecule and the interactions of multiple molecules, which could assist us in better understanding the experimental phenomena. The simulation results indicated that the R₂LS molecule has better flexibility, which was favorable for interaction with the cell membrane. And it could generate tight integration in self-assembly while R₁LS and R₃LS assemblies have a large molecular interval, which led to a controllable release of cargos for R₂LS in a reductive environment. In summary, the generation of the dendrimer in lipopeptides is vital for the gene transfer effect. For optimization, it is necessary to study the structure–function relationship, and molecular dynamic simulation is an effective strategy for screening the molecular structure and even for predicting experimental results.