Since single-point mutant perturbation has been used to probe protein folding mechanisms in experiments, the ϕ-value has become a critical parameter to infer the transition state (TS) for two-state proteins. Experimentally, large scale analysis has shown a nearly single uniform ϕ-value with normally distributed error from 24 different proteins; moreover, in zero stability conditions, the intrinsic variable ϕ⁰ is around 0.36. To explore how and to what extent theoretical models can capture experimental phenomena, we here use structure-based explicit chain coarse-grained models to investigate the influence of single-point mutant perturbation on protein folding for single domain two-state proteins. Our results indicate that uniform, additive contact energetic interactions cannot predict experimental Brønsted plots well. Those points deviate largely from the main data sets in Brønsted plots, are mostly hydrophobic, and are located in N- and C-terminal contacting regions. Heterogenous contact energy, which is dependent on sequence separation, can narrow the point dispersion in a Brønsted plot. Moreover, we demonstrate that combining many-body interactions with heterogeneous native contact energy can present mean ϕ-values consistent with experimental findings, with a comparable distributed error. This indicates that for more accurate elucidation of protein folding mechanisms by residue-level structure-based models, these elements should be considered.