Optibrium

A red glow behind a region of a molecule indicates that this region has a positive influence on the property value of the molecule in numerical terms (i.e. the value is increased). A blue glow indicates that the region has a negative influence on the property value, again in numerical terms. A green glow indicates that the region has no significant overall effect on the property. Another way to look at this is that if a ‘blue’ functional group is removed from a molecule or substituted with a ‘green’ or ‘red’ functionality, we would expect the property value to increase (for example, see the change from molecule (a) - below - to molecule (b) in which the piperazine group has been changed to a piperidine). Alternatively, if a ‘red’ functional group is removed or replaced with a ‘green’ or ‘blue’ functionality, we would expect the property value to decrease (for example, see the change from molecule (a) above to molecule (c) in which the thiophene methyl group has been removed).

It is important to note that the colours indicate changes in property value and not whether this may be good or bad for any given property, as this will depend on the objectives of the drug discovery project.

The Glowing Molecule visualisation shows an analysis of the strongest structural influences on the property of each individual molecule. More specifically, it assesses the sensitivity of the property to the presence of each atom or functional group. For a linear model a functional group would have the same influence, irrespective of the molecule in which it occurred. However, the link between structure and property is rarely linear, as the influence of a functional group will depend on the structural characteristics of the entire molecule. Therefore, the glow for a specific functional group may vary from molecule to molecule depending on the presence of other functionalities.

Occasionally, a molecule may glow completely green, even if the corresponding property value is high or low. In this case, the glow indicates that the molecule’s property value is not strongly influenced by any particular functional groups; the extreme property value is caused by the combined influence of all regions of the molecule. This indicates that small a modification to the molecule is unlikely to significantly affect the property and a more dramatic change will be necessary to change the property.

All models in the ADME QSAR module produce a Glowing Molecule output. Furthermore, models built with the Auto-Modeler using StarDrop's internal or SMARTS-defined descriptors will produce glowing molecules.

It is relatively straightforward to produce a similar visualization to Glowing Molecule for simple linear models (where the property is calculated via a weighted sum of contributions) and some other packages can generate these. However, in practice, properties are rarely linearly related to the structural characteristics of a molecule and modern modeling techniques such as artificial neural networks, Gaussian Processes and genetic programming have been used to model non-linear relationships. These methods are commonly described as ‘black box’ because understanding and visualising the non-linear models these generate has been impossible. The unique algorithm underlying Glowing Molecule enables this information to be extracted and visualised in an intuitive way to help interpret the structure-activity relationships captured in these models.