Ten Years of the MIA-QSAR Strategy: Historical Development and Applications

Theoretical Structure Of The Mia-Qsar Descriptors

The MIA-QSAR descriptors are in fact pixels of chemical structural images (i.e. the image pixels are considered as the descriptors) and are obtained as follows. For a given set of n chemical structures, with a common substructure or moiety (i.e. basic molecular scaffold), these are drawn on an m × l (same size) canvas and saved as images, separately. Subsequently, the chemical images are aligned with respect to a common coordinate forming a 3-way n × m × l Multi Variate Image (MVI). Later, the MVI is unfolded to form a 2-way n × (m × l) data matrix; with n rows and m × l columns (see Figure 1 for illustration). Since pixels can be described numerically, their coordinates in an image give rise to chemical drawings (chemical structures with different substituents in varying positions) and, therefore, distinct data matrices for each compound are obtained, which explain the variance in the properties block (dependent variables).

Figure 1.

Workflow followed in the generation of the MIA-QSAR descriptors