Estimation of physical properties of amino acids by group-contribution method

- Prediction of normal melting point, water solubility and octanol-water partition coefficient of amino acids.
- Group contribution approach to develop the property models.
- Provides uncertainty estimates for the modeled properties.
- Provides modelling details together with model parameters.

In this paper, we present group-contribution (GC) based property models for the estimation of physical properties of amino acids using their molecular structural information. The physical properties modelled in this work are normal melting point (Tm), aqueous solubility (Ws), and octanol/water partition coefficient (Kow) of amino acids. The developed GC-models are based on the published GC-method by Marrero and Gani (2001) with inclusion of new structural parameters (groups and molecular weight of compounds). The main objective of introducing these new structural parameters in the GC-model is to provide additional structural information for amino acids having large and complex structures and thereby improve the predictions of physical properties of amino acids. The group-contribution values were calculated by regression analysis using a data-set of 239 values for Tm, 211 values for Ws, and 335 values for Kow. Compared to other currently used GC-models, the developed models make significant improvements in accuracy with average absolute error of 10.8 K for Tm and logarithm-unit average absolute errors of 0.16 for Kow and 0.19 for Ws.