The Absolv Approach

Bromocriptine, erythromycin and terfenadine had to be omitted from this correlation because Absolv does not calculate the water solubilities of compounds with McGowan volumes, V, greater than 4.0. This is because the solubility training set used in Absolv, Eq. (6), does not contain compounds with V values greater than 4 hence predictive power for such compounds are not reliable. 

The coefficients of descriptors A, AB and S are not statistically significant, but the terms are left in the equation for comparison with Abraham and Le s [8] water-solubility equation, modified here to conform to the mathematical conventions used in this chapter  

Clearly, Eq. (5) and Eq. (6) are not exactly the same. In Abraham and Le s equation [8], the H-bond and volume terms have much larger coefficients than the corresponding ones in Eq. (5). This helps to explain why Eq. (6) did not predict the solubilities of the 58 compounds well. 

Nevertheless, Absolv gave us five additional parameters to work with. What is more, we used ALOGP, the predicted octanol/water partition coefficient also calculated by Absolv. We combined these descriptors with CLOGP, CMR and those from HYBOT. Forward step-wise regression gave Eq. (7)  

Comparing these results to those of Eq. (2), we find that they are nearly as good in terms of s and R2 and superior in Q2. 

---