Cheminformatics is an exciting field that combines chemistry with informatics by collecting, storing, analysing and manipulating chemical data. Cheminformatics originally emerged to aid and inform drug discovery, however it has now grown to involve important areas of biology, chemistry and biochemistry, with many similarities to the field of bioinformatics.

In this project we will first try to replicate, and then improve upon, research by John S. Delaney [1], who used linear regression to predict the aqueous solubility (which he calls ESOL values) of a range of molecules. Delaney demonstrates that you can predict the solubility to a reasonable accuracy using 4 easily observable parameters: molecular weight, number of rotatable bonds, aromatic proportion and the octane-water partition coefficient.

Aqueous solubility is one of the key physical properties of interest to a medicinal chemist as it affects the uptake/distribution of biologically active compounds within the body.

Finally, I would like to thank Prof. Jan H. Jensen [2] for the idea for this notebook and for his excellent tutorials on the use of machine learning within chemistry.

1. ESOL: Estimating Aqueous Solubility Directly from Molecular Structure, J. S. Delany, J. Chem. Inf. Comput. Sci. 2004, 44, 1000-1005
2. Machine Learning Basics by Jan H. Jensen