logS measures aqueous solubility as the base-ten logarithm of molar solubility. Increasing logS by one means tenfold greater solubility. section 4.1.2
Use logS to interpret solubility and prepare prediction data:
- Values around -6 to -3 correspond approximately to 1 μM to 1 mM, a common window for drug-like compounds. section 4.1.2
- Values below -6 indicate poor solubility and can signal bioavailability challenges. section 4.1.2
- Inspect extreme values before modeling: the dataset spans -12 to 1.97, and outliers may impair learning. section 4.1.2
- Flag candidates statistically with the IQR rule or remove implausible compounds using chemical domain knowledge. section 4.1.2
Figure 4.1 Comparison of solubility distributions for train and validation data sets.

Compare training and validation solubility distributions to spot shifts and unusually broad ranges. section 4.1.2
