APPLICATION OF ANT COLONY OPTIMIZATION FOR SELECTION OF VARIABLES IN SECOND ORDER DATA APPLIED TO DETERMINATION OF FLUOROQUINOLONE ANTIBIOTICS IN HUMAN SERUM

SABIN, G.M.; POPPI, R.J.; OLIVIERI, A. C.

Budapest

Congreso; XIII Conferencia en Quimiometría en Química Analítica; 2012

Hungarian Chemical Society

A new variable selection methodology based on ant colony optimization (ACO) [1] was applied in a three-way fluorescence data [2]. The algorithm, written in Matlab, is based on the concept of cooperative pheromone accumulation to minimize the standard error of validation (SEV) by means of an objetive function using UPLS models. In this work, three fluoroquinolones (norfloxacin, enoxacin and ofloxacin) were prepared by a design of experiments (DOE) resulting 15 calibration and 8 external validation samples with total of 23 emission-excitation fluorescence matrices (EEMs). As a way for improve efficiency of ACO algorithm, a full factorial 35 design, using MODDE 9.0, was performed to choose ACO parameters to be used in SEV minimization. Thus, the following results were achieved for norfloxacin, enoxacin and ofloxacin, respectively: number of generations: 300, 500 and 100; total ants: 20, 10 and 30; width of variable block: 4, 3 and 4; number of blocks in excitation dimention: 5, 5 and 8; number of blocks in emission dimention: 2, 2 and 2. The results show SEVs and respective range of concentration used in DOE: norfloxacin: 0,3 ug L-1 (range: 0-25 ug L-1); enoxacin: 4,0 ug L-1 (range: 0-300 ug L-1); and ofloxacin: 0,4 ug L-1 (range: 0-80 ug L-1). These results are better understood by graphical representations of the selected regions on each fluoroquinolones profile, Figure 1. Figure 1 – Ant Colony Optimization. a) EEM of all samples of the DOE; b) selected variables of norfloxacin profile*; c) selected variables of enoxacin profile*; d) selected variables of ofloxacin profile*.