The purpose of the olfactory system is to encode and classify odorants. Hence, its circuits have likely evolved to cope with this task in an efficient, quasi-optimal manner. In this chapter the authors present a three-step approach that emulate neurocomputational principles of the olfactory system to encode, transform and classify chemical data. In the first step, the original chemical stimulus space is encoded by virtual receptors. In the second step, the signals from these receptors are decorrelated by correlation-dependent lateral inhibition. The third step mimics olfactory scent perception by a machine learning classifier. The authors observed that the accuracy of scent prediction is significantly improved by decorrelation in the second stage. Moreover, they found that although the data transformation they propose is suited for dimensionality reduction, it is more robust against overdetermined data than principal component scores. The authors successfully used our method to predict bioactivity of drug-like compounds, demonstrating that it can provide an effective means to connect chemical space with biological activity.
TopBackground
Figure 1 outlines the basic architecture of an olfactory system. “Chemical space” consists of the multitude of odorant molecules which float around us in the air. These molecules activate an array of receptor neurons. Subgroups of receptor neurons are distinguished into classes according to the particular olfactory receptor protein they present on the membrane (indicated by different shades of gray in Figure 1). Since this membrane-bound receptor protein determines the ligand selectivity of the receptor neuron, neurons of one class respond to the same odorants.
Figure 1. Schematic of the basic architecture of the olfactory system in insects and mammals
The number of receptor neuron classes varies greatly between species and exhibits weak correlation with olfactory capability. The fruit fly Drosophila melanogaster possesses about 60 different functional receptor genes (Vosshall, Wong, & Axel, 2000), the honeybee Apis mellifera has about 160 (Robertson & Wanner, 2006). Humans are believed to possess approximately 250 different receptor classes (Glusman, Yanai, Rubin, & Lancet, 2001, Zozulya, Echeverri, & Nguyen, 2001), while mice have about 1000 (Zhang & Firestein, 2002) and dogs around 1200 (Olender et al., 2004).
Olfactory receptor neurons typically exhibit broad ligand selectivity (Araneda, Kini, & Firestein, 2000, Bruyne, Foster, & Carlson, 2001, Mori, Takahashi, Igarashi, & Yamaguchi, 2006, Hallem & Carlson, 2006). A receptor neuron of a given class is therefore activated by many different odorants, as well as an odorant may activate many different classes of receptor neurons. As a consequence, the identity of the odor stimulus is encoded in the combinatorial activity pattern of all receptor neurons rather than having a dedicated receptor for each odorant.
Receptor neurons form the first neuronal layer in the olfactory system. They project to a secondary layer, the antennal lobe in insects, or the olfactory bulb in mammals, respectively. This layer is organized into so-called glomeruli — small compartments where receptor neurons make contacts with their downstream partners (projection neurons in insects; mitral cells in mammals). Receptor neurons from one class (i.e. with the same ligand selectivity) converge onto the same glomerulus. Hence, an odorant evokes a specific pattern of activity on the glomeruli.