شناسایی حالت چرخشی داخلی ممانعت کننده برای گونه های پیچیده شیمیایی: رویکرد داده کاوی با مدل رگرسیون چندگانه لجستیک

Thermodynamic properties are essential to understand and describe many chemical/biological processes in the real environment. To obtain correct thermodynamic data of chemical species for a wide range of temperatures, a rigorous Hindered Internal Rotation (HIR) treatment must be considered. Such a treatment requires detailed information about the internal rotation (i.e., rotational axis, group, frequency, symmetry and hindrance potential). However, it is very tedious, even prone-to-error, for chemists to prepare the input parameters for such a treatment. Among the HIR parameters, the rotational frequency (or mode) is the most difficult element due to the complex molecular structure and mixing vibrational modes of chemical species. Recently, a rule-based framework has been proposed to help chemists with this tedious process (Le et al., Comput. Theor. Chem., 2017, 61). This approach has been demonstrated to work well for simple species; however, it still lacked the ability to handle more complex cases. Therefore, in this study, a data mining approach is proposed to overcome the challenges of the previous algorithm. Within this framework, the HIR pattern was found using the features extracted from existing data provided by chemists. More specifically, multivariate logistic regression was implemented to analyze the chemical data to better predict the rotational frequency (mode) of chemical species as well as to highlight the effect of each attribute of the rotation. The experimental results were demonstrated to be more accurate than the previous study in terms of both accuracy and completeness. It also gives meaningful insights into the HIR itself. The proposed approach of this research will be integrated into MSMC-GUI (https://sites.google.com/site/msmccode/manual/gui-1) to provide chemists with both an interactive and robust tool to prepare the data for their thermodynamic calculations on-the-fly.