We study the resonant spin-flavor (RSF) conversion of supernova neutrinos, which is induced by the interaction between the nonzero neutrino magnetic moment and the supernova magnetic fields, and its dependence on presupernova models. As the presupernova models, we adopt the latest ones by Woosley, Heger, and Weaver, and, further, models with both solar and zero metallicity are investigated. Since the $(1-2Y_e)$ profile of the new presupernova models, which is responsible for the RSF conversion, suddenly drops at the resonance region, the completely adiabatic RSF conversion is not realized, even if $\mu_\nu B_0= (10^{-12}\mu_B)(10^{10} G)$, where $B_0$ is the strength of the magnetic field at the surface of the iron core. In particular for the model with zero metallicity, the conversion is highly nonadiabatic in the high energy region, reflecting the $(1-2Y_e)$ profile of the model. In calculating the flavor conversion, we find that the shock wave propagation, which changes density profiles drastically, is a much more severe problem than it is for the pure Mikheyev-Smirnov-Wolfenstein (MSW) conversion case. This is because the RSF effect occurs at a far deeper region than the MSW effect. To avoid the uncertainty concerning the shock propagation, we restrict our discussion to 0.5 s after the core bounce (and for more conservative discussion, 0.25 s), during which the shock wave is not expected to affect the RSF region. We also evaluate the energy spectrum at the Super-Kamiokande detector for various models using the calculated conversion probabilities, and find that it is very difficult to obtain useful information on the supernova metallicities and magnetic fields or on the neutrino magnetic moment from the supernova neutrino observation. Future prospects are also discussed.