Mira Zečević

Application of neural networks for response surface modeling in HPLC optimization

The usefulness of artificial neural networks for response surface modeling in HPLC optimization is compared with multiple regression methods. The results show that neural networks offer promising possibilities in HPLC method development. The predicted capacity factors of analytes were better to those obtained with multiple regression method.

Prediction of drug transfer into human milk from theoretically derived descriptors

Abstract The goal of this study was to develop a genetic neural network (GNN) model to predict the degree of drug transfer into breast milk, depending on the molecular structure descriptors, and to compare it with the current model. A supervised network with back-propagation learning rule and multilayer perceptron (MLP) architecture was used to correlate activity with descriptors that were preselected by a genetic algorithm. The set of 60 drug compounds and their experimentally derived M / P values used in this study were gathered from literature. A total of 61 calculated structural features including constitutional, topological, chemical, geometrical and quantum chemical descriptors were generated for each of the 60 compounds. The M / P values were used as the ANNs output and calculated molecular descriptors as the inputs. The best GNN model with 26 input descriptors is presented, and the chemical significance of the chosen descriptors is discussed. Strong correlation of predicted versus experimentally derived M / P values ( R 2 >0.96) for the best ANN model (26-5-5-1) confirms that there is a link between structure and M / P values. The strength of the link is measured by the quality of the external prediction set. With the RMS error of 0.425 and a good visual plot, the external prediction set ensures the quality of the model. Unlike previously reported models, the GNN model described here does not require experimental parameters and could potentially provide useful prediction of M / P ratio of new potential drugs and reduce the need for actual compound synthesis and experimental M / P ratio determination.

Spectrophotometric study of diclofenac-Fe(III) complex

A multifactor optimisation technique is successfully applied to develop a new spectrophotometric method in which diclofenac sodium is analysed and determined as its Fe(III) complex. The effect of simultaneously varying the pH, ionic strength and concentration of colour reagents in the reaction mixture were studied. A four-variable two-level factorial design was used to investigate the significance of each variable and interactions between them. A response surface design was used to optimise complex formation and extraction. It was established that diclofenac reacts with Fe(III) chloride, in the presence of ammonium thiocyanate, in the pH range 4.2-6.5, forming a red chloroform extractable (2:1) complex with maximum absorbance at 481 nm. By applying the methods of Sommer and Job involving non-equimolar solutions the conditional stability constant of the complex, at the optimum pH of 6.0 and an ionic strength mu = 0.19M, was found to be 10(6.4). Good agreement with Beers law was found for diclofenac concentrations up to mmol 1(-1). The nominal percent recovery of diclofenac was 98.8% (n = 10). The lower limit of sensitivity of the method was found to be 14.7 micrograms ml(-1).

Application of artificial neural networks in HPLC method development

The use of artificial neural networks (ANNs) for response surface modelling in HPLC method development for amiloride and methychlothiazide separation is reported. The independent input variables were pH and methanol percentage in mobile phase. The outputs were capacity factors. The results were compared with a statistical method (multiple nonlinear regression analysis). Networks were able to predict the experimental responses more accurately than the regression analysis.

Application of experimental design in optimization of solid phase extraction of mycophenolic acid and mycophenolic acid glucuronide from human urine and plasma and SPE-RP-HPLC method validation

The aim of this study was to develop and optimize a solid phase extraction (SPE) procedure for purification of mycophenolic acid (MPA) and its metabolite mycophenolic acid glucuronide (MPAG) in biological samples. During optimization process chemometric approach was applied. First, in screening experiments fractional factorial design (FFD) was used for selecting the variables which affected the extraction procedure. The ionic strength of the phosphate buffer in the washing step and the percentage of acetonitrile in the elution step were statistically significant for the recovery of MPAG while the percentage of acetonitrile and pH of the washing solution were statistically significant for that of MPA. Afterwards, the significant variables were optimized using central composite design (CCD). The developed SPE method included phosphate buffer (pH 2.4; 0.056 M) in the washing step, and the mixture of acetonitrile and phosphate buffer of which pH was adjusted to 2.4 (70:30, v/v) in the elution step. The investigation was applied to both urine and plasma and the nature of biological matrix appeared to be of no importance. The extraction from both matrixes showed good repeatability with relative standard deviations up to 6% for MPAG and 8% for MPA, and recovery around 100% for both substances. Furthermore, new SPE-RP-HPLC method for determination of MPA and MPAG in both humane urine and plasma has been validated. The great advantage of this method is the chromatographic run of only 3 min.

Validation of a high-performance liquid chromatography method for the determination of pancuronium in Pavulon injections.

A new high-performance liquid chromatography (HPLC) method was developed for the quality control of pancuronium bromide and its degradation products. The HPLC method used a 5-microm Supelcogel ODP-50 (150x4 cm) column with acetonitrile-CH3OH-water-F3CCOOH (20.5:74.9:0.1, v/v) as the mobile phase (pH value 2.0 adjusted with trifluoroacetic acid) at a flow-rate 0.8 ml/min and UV detection at 210 nm. The Beers law plots were found to be linear over the concentration range 0.4-1.2 mg/ml of pancuronium bromide and 0.04-0.08 mg/ml of desacetyl degradation products (R2=0.9995). The RSD of the peak areas was 1.09% and the recovery was 102.43%. The RSD value shows good precision, acceptable accuracy and reproducibility of the new method for the determination of pancuronium bromide in presence of its desacetyl degradation products. The method is rapid and sensitive enough to be used for Pavulon injection analysis.

Use of ANN modelling in structure--retention relationships of diuretics in RP-HPLC.

Structure retention relationship study, conducted by RP HPLC, was used to investigate physical chemical parameters related to the RP retention times of amiloride, hydrochlorothiazide and methyldopa in order to predict the separation of amiloride and methylclothiazide from Lometazid tablets. Retention data were obtained with an ODS column using a mobile phase methanol water (pH adjusted with phosphoric acid). Physical chemical properties were calculated directly from the molecular structure. Artificial neural networks (ANNs) were used to correlate chromatograms retention times with mobile phase composition and pH, and with physical chemical properties of amiloride, hydrochlorothiazide and methyldopa and to predict separation of amiloride and methylclothiazide from Lometazid tablets. Sensitivity analysis was performed to interpret the meaning of the descriptors included in the models. Results confirmed the dominant role of the polar modifier in such chromatographic systems. Within a series of solutes chromatographed under identical conditions, the retention parameters could be approximated by a non-linear combination of logP, logD, pKa, surface tension, parachor, molar volume and to minor extend by polarisability, reetractivity index and density. This study has demonstrated that the use ANNs techniques can result in much more efficient use of experimental information. As HPLC is the most popular analytical technique, improvements in HPLC methods development can yield significant gains in the overall analytical effort. The ANNs extension presented could be the method of choice in some advanced research settings and serves as an indication of the broad potential of neural networks in chromatography analysis.

Study of forced degradation behavior of Eletriptan hydrobromide by LC and LC–MS and development of stability-indicating method

The objective of the present study was to report the stability profile of novel antimigrain drug Eletriptan hydrobromide based on the information obtained from forced degradation studies. The drug was subjected to acid (0.1-1 mol L(-1) HCl), neutral and base (0.1-1 mol L(-1) NaOH) hydrolysis and to oxidative decomposition (3-15% (v/v) H(2)O(2)). Photolysis and thermo degradation at 75 degrees C were carried out in methanol solution and in solid state with both Eletriptan hydrobromide bulk drug and the tablet formulation. The products formed under different stress conditions were investigated by LC and LC-MS. The experimental conditions for LC were chosen by employing experimental design and multicriteria decision making methodology. These powerful tools enabled the accomplishment of satisfactory resolution with the shortest possible analysis time. Analytes were separated on a C(18) column (XTerra, 150 mm x 3.9 mm, 5 microm) with the mobile phase composed of methanol-water solution of TEA (pH 6.52, 1%, v/v) (30:70, v/v) pumped at 1 mL min(-1) flow rate. The column temperature was set at 50 degrees C and the detection at 225 nm using DAD detector. The LC method was suitably modified for LC-MS analysis which was further used to characterize the arisen degradation products. The possible degradation pathway was outlined based on the results. The drug appeared to be instable towards every stress condition but oxidation. The stability was not jeopardized even under more exaggerated conditions such as increased temperature of the solutions to 75 degrees C, increased strength of acid/alkali solutions and prolonged testing period. Validation of the LC-DAD method was carried out in accordance with ICH guideline. The method met all required criteria and was applied when testing the commercially available tablets.

Multicriteria optimization methodology in development of HPLC separation of mycophenolic acid and mycophenolic acid glucuronide in human urine and plasma.

Multicriteria optimization methodology was applied for development of isocratic reversed-phased HPLC method for simultaneous determination of mycophenolic acid (MPA) and mycophenolic acid glucuronide (MPAG) in human urine and plasma. In the first stage of method development, pH value of the water phase, percentage of acetonitrile, temperature of the column and flow rate of the mobile phase were investigated using fractional factorial design. Afterwards, the optimal conditions were found employing central composite design and Derringers desirability function. Two goals were considered, the retention factor of the MPAG to be in the range, between 0.8 and 1.118 which allowed well separation of MPAG from the urine and plasma peaks, and the shortest possible total analysis run time. Then, the obtained sigmoid functions were used to transform the optimization criteria into the desirability values. The satisfactory chromatographic conditions were obtained with mobile phase consisted of acetonitrile-phosphate buffer (pH 2.4; 0.04 M KH(2)PO(4)) (28:72, v/v). The separation was performed on C(18) Chromolith column (100 mm x 4.6 mm) with flow rate of 5 mL/min, the temperature of the column was 25 degrees C and the chosen wavelength for simultaneous determination of MPA and MPAG was 215 nm. The MPAG eluted at 0.552 min and the duration of run was 3.092 min. Afterwards, the method was subjected to validation. Linearity was observed over the concentration range of 1-50 microg/mL for MPA and 1-500 microg/mL for MPAG in urine and 1-60 microg/mL for MPA and 1-70 microg/mL for MPAG in plasma matrix. The method showed intra-day and inter-day precision with relative standard deviation lower then 5% and accuracy as recovery (%) between 100+/-5%.

Quantitative structure-retention relationships of azole antifungal agents in reversed-phase high performance liquid chromatography.

Artificial neural network (ANN) is a learning system based on a computational technique which can simulate the neurological processing ability of the human brain. It was employed for building of the quantitative structure-retention relationships (QSRRs) model of antifungal agents-imidazoles or triazoles by structure. Computed molecular descriptors together with the percentage of acetonitrile in mobile phase (v/v) and buffer pH, being the most influential HPLC factors, were used as network inputs, giving the retention factor as model output. The multilayer perceptron network with a 9-5-1 topology was trained by using the back propagation algorithm. Good correlation between experimentally obtained data and ones predicted by using QSRR-ANN on previously unseen data sets indicates good predictive ability of the model.