Experimental data on optical resonance spectra of whispering gallery modes of dielectric microspheres in antibiotic
solutions under varied in wide range concentration are represented. Optical resonance was demonstrated could be
detected at a laser power of less than 1 microwatt. Several antibiotics of different generations: Amoxicillin,
Azithromycin, Cephazolin, Chloramphenicol, Levofloxacin, Lincomicin Benzylpenicillin, Riphampicon both in deionized
water and physiological solution had been used for measurements. Both spectral shift and the structure of
resonance spectra were of specific interest in this investigation. Drag identification has been performed by developed
multilayer perceptron network. The network topology was designed included: a number of the hidden layers of
multilayered perceptron, a number of neurons in each of layers, a method of training of a neural network, activation
functions of layers, type and size of a deviation of the received values from required values. For a network training the
method of the back propagation error in various modifications has been used. Input vectors correspond to 6 classes of
biological substances under investigation. The result of classification was considered as positive when each of the region,
representing a certain substance in a space: relative spectral shift of an optical resonance maxima - relative efficiency of
excitation of WGM, was singly connected.
It was demonstrated that the approach described in the paper can be a promising platform for the development of
sensitive, lab-on-chip type sensors that can be used as an express diagnostic tools for different drugs and instrumentation
for proteomics, genomics, drug discovery, and membrane studies.