Multivariate calibration methods applied to the monitoring of the enzymatic synthesis of amipicilin

An integrated semi-continuous rector, with simultaneous crystallization of the antibiotic, is the most promising alternative for the enzymatic synthesis of semi-synthetic penicillins. Closed-loop optimization of the reactor operation requires the measurement of its state variables. An efficient method for on-line monitoring of these variables, i.e., the species concentrations, which is consistent with the dynamics of the process, is presented here. Multivariate calibration has been used to reduce times of analysis in multicomponent systems, and this is the approach of this work. Three multivariate calibration methods are compared for prediction of the concentrations of the components present in the enzymatic synthesis of ampicillin. UV spectra and their first-order derivatives were used as input data for the models. Results showed that all methods (SPA, PCR and PLS) had similar performances for this system. The use of first-order derivatives of the spectra did not improve substantially the performance of the methods. Bench assays of synthesis of ampicillin using immobilized penicillin G acylase (PGA) as catalyst were carried out to validate the methodology. Its results were compared with HPLC analyses. The technique showed to be very accurate for monitoring the concentrations of ampicillin (desired product) and 6-aminopenicillanic acid (6-APA, most expensive substrate), with precision similar to those obtained by HPLC. Prior information of the initial reactor state may be used to improve the estimation of less sensitive components (phenylglycine methyl ester, PGME and phenylglycine, PG). Modifications on the biocatalyst must be made in order to use the proposed automatic sampler when crystallization of products occurs.