Solubility has become one of the key physicochemical screens at early stages of the drug development process. Solubility prediction through Quantitative Structure-Property Relationships (QSPR) modeling is a growing area of modern pharmaceutical research, being compatible with both High Throughput Screening technologies and limited compound availability characteristic of early stages of drug development. We resort to the QSPR theory for analyzing the aqueous solubility exhibited by 145 diverse drug-like organic compounds (0.781 being the average Tanimoto distances between all possible pairs of compounds in the training set). An accurate and generally applicable model is derived, consisting on a linear regression equation that involves three DRAGON molecular descriptors selected from more than a thousand available. Alternatively, we apply the linear QSPR to other 21 commonly employed validation compounds, leading to solubility estimations that compare fairly well with the performance achieved by previously reported Group Contribution Methods.