[Prediction and estimation on molar response values of alkane using molecular subgraph]

Area of Science:

• Computational Chemistry
• Cheminformatics
• Quantitative Structure-Property Relationship (QSPR)

Background:

• Molar response values are crucial for quantitative analysis of alkanes using techniques like Gas Chromatography.
• Accurate prediction of these values is essential for chemical analysis and molecular design.
• Existing methods may require extensive experimental data or complex calculations.

Purpose of the Study:

• To develop a novel and efficient method for characterizing alkane molecular structures.
• To establish a strong correlation between the proposed molecular coding and molar response values.
• To validate the predictive capability of the new method using advanced statistical and machine learning techniques.

Main Methods:

• A new coding scheme based on topographical subgraphs of alkanes was developed.
• This coding was used to represent molecular structures.
• Multiple linear regression (MLR) was employed to correlate the coding with molar response values.
• Back-propagation neural network (BPNN) and leave-one-out cross-validation were used for model validation.

Main Results:

• A highly significant correlation (r=0.9983 for FID, r=0.9963 for TCD) was found between the topographical subgraph coding and molar response values.
• MLR models provided excellent fits for predicting molar response values on both Flame Ionization Detector (FID) and Thermal Conductivity Detector (TCD).
• Cross-validation using BPNN yielded a high correlation coefficient of 0.989, confirming the model's robustness.

Conclusions:

• The proposed topographical subgraph coding is an effective descriptor for alkane molecular structures.
• This method offers a reliable and accurate approach for predicting molar response values.
• The findings suggest potential applications in quantitative structure-property relationship studies and chemical analysis.