The partial least squares(PLS) method was employed to establish a quantitative ion characteristics-activity relationship(QICAR) model for metal ion toxicity(EC50 of 15 metal ions). The ion characteristics included AN(the atomic number), ΔIP(the change in ionization potential, eV), Xm(the electronegativity, eV), AW(the atomic weight), Xm2r(the covalent index), ΔE0(the absolute difference between electrochemical potential of the ion and that of its first stable reduced state, eV),|lgKOH|(the absolute value of the lg of the first hydrolysis constant), AR(the atomic radius, nm), AR/AW(the ratio between atomic radius and atomic weight) and σp(the softness index) selected based on relative correlation analysis. The simulated and tested(with the other four metals) efficiency coefficients of the model are 0.88 and 0.96, respectively. The information revealed from the QICAR model indicates that the value of the metal ion toxicity was positively correlated with variables AN, ΔIP, Xm, AW and Xm2r; negatively correlated with variables ΔE0, |lgKOH|, AR/AW, AR and σp, and ion characteristics ΔE0, Xm, σp and Xm2r were found to contribute more to the toxicity of metal ions via the accurate analysis method provided by PLS. The model could be used to predict the toxicity of the target metals and preliminary to assess combined pollution and environmental risk for heavy metals in the environments.