Abstract
In continuing with the research program initiated by Llored and Harré of exploring the part/whole discourses of chemistry, we analyse Quantitative Structure-Activity Relationships studies, which are widespread approaches for modeling substances’ properties. The study is carried out by analyzing a particular QSAR model, and it is found that different mereologies are needed: from those regarding bulk substances as wholes and molecular entities as parts and to mereologies where the wholes are molecules whose parts are atoms, structured subsets of atoms, nuclei and electronic densities. We suggest a relationship between successful QSAR models and a deep understanding of the mereologies used and the ways they are intertwined. We note that QSAR modelers prefer the mereology of substance-molecule and then discuss how that is related to simplicity and computational capacity. Historical questions are opened, e.g. how the mereologies of substances have changed over time? and why they are mostly oriented toward organic chemistry?