[DSRP Evidence](https://dsrpevidence.org/)

# Mereology of Quantitative Structure-Activity Relationships Models

## Details

**Authors** Restrepo, Guillermo and Harré, Rom

**Year** 2015

**Publisher** HYLE: International Journal for Philosophy of Chemistry, 21(1), 19-38

**Kind of work** article

**Discipline** Chemistry

**Secondary disciplines** Philosophy

**Applied** false

[Read it](https://hyle.org/journal/issues/21-1/restrepo-harre.htm)

## In authors' words

### What they set out to do (purpose)

To analyze the mereological structure of quantitative structure-activity relationship (QSAR) models in chemistry.

### Who or what was studied (sample)

QSAR modeling practice, analyzed conceptually.

### How they did it (methods)

Conceptual analysis. Each part-whole structure used in QSAR is checked against a relaxed parthood relation for reflexivity, antisymmetry, transitivity, company, supplementation and atomicity.

### What they found (results)

QSAR modeling relies on several mereologies rather than one. The whole is first a bulk substance whose parts are molecules. Molecules then become wholes in descriptor calculations, with atoms, substructures, nuclei, electron densities or graph vertices and edges as parts. The simple substance-molecule mereology is atomic and loses information, while richer ones such as the molecular graph satisfy more of the formal properties. Whether a part can stand for its whole depends on context such as pH, temperature or biological conditions. Modelers favor the simple mereology, which the authors attribute to a trade-off between simplicity and computational capacity.

## Commentary

### In short

The authors show that chemical models of structure and biological activity do not rest on a single way of dividing a substance into parts. They trace how the choice of whole and parts shifts across the modeling workflow and what each choice gains and loses.

### Response

The paper bears most directly on the systems structure: the same material is treated as a whole at one stage and as a part at the next, and the choice of decomposition shapes what a model can predict. The dependence of part-for-whole substitution on conditions such as pH and temperature, and the authors' note that modeling choices depend on researcher experience and negotiation, point to a perspective-dependence of the decomposition, though the authors do not frame it that way. The fit for perspectives is therefore partial. The paper is conceptual and offers no test of whether the choice of mereology changes model performance.

**Patterns it shows** S, P

Shows the structure is there

**Added** 2026-10-19

**How to cite this** Restrepo, Guillermo and Harré, Rom (2015). Mereology of Quantitative Structure-Activity Relationships Models. HYLE: International Journal for Philosophy of Chemistry, 21(1), 19-38.