Matthews correlation coefficient

A principal component analysis (PCA) using simple molecular descriptors showed that the training and test sets overlapped. Focusing on compounds with a Tanimoto similarity greater than 0.7 resulted in a test set of 28 compounds, which had Matthews correlation coefficient and concordance statistics that... 

Matthews correlation index = Pearson similarity coefficient —> classification parameters > maximal binding energy —> scoring functions (0 average binding energy)... 

Similarity coefficients for binary variables are also used as —> classification parameters for two-dass problems among these, the most used is the Pearson coeffident (Table S9), which is also known as Matthews correlation index. 

SUBSTRUCT prediction CNS-, and all molecules with a score >0.5 are classified as CNS-i-. The correlation coefficient (CC) according to Matthews [55] multiplied by 100, and the prediction accuracy (PA) for compounds within specific scoring interval is included. The SUBSTRUCT classification (right) is analogous to the ANN classification [40]. 

Bockris, Khan, and Matthews pointed out that if one plots the value of calculated on the solvent fluctuation theory for a number of redox reactions in solution, against the corresponding value, calculated from experiment, with the transmission coefficient assumed to be unity, one obtains no correlation, except that the values theoretically predicted are generally below those observed experimentally. 

Molecular simulation using realistic force fields provides a reliable prediction of diffusion coefficient (D). However, it is computationally expensive and cannot be used for routine engineering calculations. An alternative approach is to develop phenomenological correlations that represent accurately experimental or molecular simulation data and can be interpolated or even extrapolated in other range of conditions. In the past, many empirical correlations have been developed with varying degree of success (Wilke and Chang (1955), Leahy-Dios and Firoozabadi (2007), Matthews et al. (1987) and Erkey et al. (1990)) [10] [13] [6]. 

The knowledge of the diffusion coefficient is crucial for the proper design of many important industrial processes. (Matthews and Akgerman, 1987) [13] and (Erkey et al., 1990) [6] developed an empirical correlation for predicting the diffusion coefficients of gases and liquid solutes in normal alkanes. The functional form of the correlation is... 

---