You are on page 1of 8

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

JCPR 2011;7 (1): 21-28 2010 Medipoeia Received 14-9-2011 Revised: 15-9-2011 Accepted: 17-9-2011

Optimization techniques in pharmaceutical industry: A Review


Ritesh N. Sharma* Shyam Sunder Pancholi

ABSTRACT Optimization techniques are abundant in pharmaceutical industry. In general, all the required information should be obtained from as few experiments as possible. Conventional techniques such as response surface models or simplex optimization are often used. With the advent of the computer in the laboratory, a new class of optimization problems arose which could not be tackled with the standard methodologies. For these search type problems, new strategies such as simulated annealing (SA), genetic algorithms (GA) and artificial neural network are applied. Although these are not guaranteed to give the optimal result, in almost all cases they are able to find very good solutions where other techniques fail completely. These methods find themselves in an intermediate position between the strong methods, and weak methods. This article provides an overview of conventional and recent optimization techniques. Keywords: simulated annealing and artificial neural network.

Ritesh N. Sharma* S. K. Patel College of Pharmaceutical Education and Research, Ganpat University, Mahesana-Gozaria Highway, Kherva Dist. Mahesana382711, Gujarat, India Shyam Sunder Pancholi Babria Institute of Pharmacy, Varnama, Vadodara-391240, Gujarat, India

1. INTRODUCTION
In order to design the best formulation it is of course possible to use a trial and error approach but this is not an effective way. Systematic optimization techniques are always preferable. These methods can be divided into sequential methods, simultaneous methods or combinations of both (De Boer et.al. 1988). Optimization refers to the art and science of allocating available resources to the best possible effect. Optimization techniques are used in industrial planning, allocation, scheduling, decision-making, etc. New optimization techniques are arriving daily, often stimulated by fascinating insights from other fields. Genetic algorithms, for example, use an anology to chromosome encoding and natural selection to evolve a good optimized solution. Certain characteristics of their architecture and the way they process information makes them superior to conventional techniques on certain class of optimization techniques. Many different optimization methods exist and they can be classified in a number of ways. The optimization techniques thus formulated can be classified as conventional and recent based on the type of objective and constraints (De Boer et.al. 1991). In pharmaceutical technology, one is often engaged with the design and analysis of formulations. A formulation of a solid dosage form consists normally of the medicinal-substance and a number of excipients (Dick et.al. 1988). In many cases, the physical properties of the formulation are determined by the physico-chemical properties of these excipients and the process of manufacturing In order to design the best formulation it is of course possible to use a trial and error approach but this is not an effective way. In the modern laboratory, the advent of complicated instruments controlled by computers has changed the nature of the problems faced by the chemist. Whereas the actual doing of the experiment used to be a major limiting factor in obtaining relevant information, and therefore warranted a large investment in time and energy to be designed and executed, in the modern laboratory experiments can be performed in a fast, reproducible way. Optimization is the search for a maximum or minimum in the value of a certain response function. For example, the yield of a chemical reaction is a function of several variables, such as concentrations of components in the mixture and physical characteristics such as temperature and pressure (Doornbos et.al. 1981).

Correspondence: Ritesh N. Sharma* S. K. Patel College of Pharmaceutical Education and Research, Ganpat University, Mahesana-Gozaria Highway, Kherva Dist. Mahesana382711, Gujarat, India E-mail:- riteshin001@yahoo.co.in

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

Such optimizations are typically performed in an iterative fashion (Figure 1). The search is started either from one or more random positions or from a set of points, picked according to some criterion.

Method Selection SA

Characteristic

Accept new solution if better, else accept with probability e-E/T

TS

Accept best of all admissible new solutions, even if it is worse than the previous solution

GA

Accept solutions according to their relative quality in the population

Generation SA Randomly pick one solution out of the

neighborhood Fig. 1. The basic iterative optimization cycle. The field of optimization is broad and has applications in all areas of chemistry and pharmacy. Naturally, many different methods have been used and in some cases even developed by chemists. In this article, an overview is how to apply optimization techniques in pharmaceuticals formulation. 2. CONVENTIONAL OPTIMIZATION METHODS Hybrid Methods Hybrid methods combine the characteristics of different optimization methods. hybrid methods are most useful when they combine the best features of their components. One approach which has been in practice is a hybrid form of a GA and SA. The GA is used to maintain a pool of trial solutions and to generate offspring; the SA governs the selection part by accepting all improving (Mackay et.al. 1998). The advantages of the GA are that inadmissible regions of the search space are handled gracefully and do not constitute a barrier that a vintage SA may find difficult to cross. The advantage of the SA is the added control that the user may exert through the cooling scheme. For the selection and generation stages, typical strategies are summarized in Table 1. Other possible hybrids are GA/TS . The main strengths of all these methods their versatility and their ability to adapt to all kinds of problems are also their weakness (Cornell J.A. 1990). Monte Carlo Simulated Annealing MCSA, also known as the Metropolis algorithm, is probably the most used optimization method in chemistry today. Although the principle of the method was used in 1953 by Metropolis it only became popular after 1983 when Kirkpatrick described its use as a general optimization scheme. The method is implemented in a large number of software packages and is the de facto standard in structure optimization problems (Hobbs et.al. 1998). TS GA Generate all solutions in neighborhood Use random based procedures such as cross-over and mutation to generate new solutions

Table 1. Selection and generation characteristics for SA, TS, and GA The principle of the method is very simple indeed. A random walk is performed in the search space (Papahadjopoulos et.al. 1995), accepting all moves that lead to a better solution, and accepting moves that lead to a worse solution with a probability e_1E=T. This is analogous to the well-known Boltzmann distribution. The only part that must be provided by the user are an evaluation function, a procedure to generate a new trial state and a cooling scheme. The algorithm is stopped when the optimal solution has been found, a predefined number of evaluations is reached, or no improvement has been obtained for a specified number of evaluations (Tricot et.al. 1984). Random Search Random search is not a strategy many people would use, unless there is no other alternative. Yet in some molecular mechanics software packages it is implemented to serve as a reference point for other optimization procedures (Press et.al. 1988). The strategy is simple: just keep on trying new candidate solutions and keep the best one(s) until the time is up. A typical example is molecular recognition seemingly small changes to an active molecule can lead to a drastic decrease in activity. It is therefore not surprising that, especially in combinatorial chemistry, the random search strategy has led to the discovery of several interesting and completely new drug leads, although also in this field more sophisticated techniques are being investigated (Barron, A. R. 1993).

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

Multivariate methods The development of modern analytical equipment provides the researcher with larg quantities of data. Extracting useful information from the collected data becomes a new challenge for the researcher. Just looking at data tables or examining one variable at a time is not enough in most cases (Schmidt et.al. 1993). PCA (Schmidt A.H. 1993 & Wolcott et.al. 2000) The objective of PCA is to describe the variation in data with a minimum of variables. Variables are often dependent on each other, PCA shows the underlying structure n the data. The principle of PCA are illustrated in two dimensions in equation1 gives the mathematical expression for the PCA model. X= 1 .x + T.P+ E (1)

A data matrix X consists of N object described by K variables. The parameter consist x determines the center of the data set, T is the score vector matrix, P is the loading matrix and the matrix E contains the residuals, the part of data matrix explained by PC model. The score values for two principle components, e.g. t 1 and t2 together span a mathematical plane to form a two dimensional model of the data. One could say the t1/t2 score plot constitute a window through which data can be viewed. Grouping and trends in data are more easily discovered in this way, outliers can be detected (Markowski W. 1996). PLS (Markowski W. 1993) PLS takes PCA one step further, as it deals with both X and Y data. X data could be variables such as reaction temperature and pH, and Y data could be responses in the form of yield. The underlying structures in X and Y data set are related to each other. Each object is represented in both X-space and Y-space. The first step is same for PLS and PCA. X = 1.x+T.P+E Y = 1.y+U.C+F (2) (3)

Fig. 2. Illustration of the SIMCA method. Around one or more classes a tolerance interval is constructed (a). New objects that fit inside the interval are said to belong to the class (b) (Chloupek et.al. 1992) Multivariate Design (Baba et.al. 1991) Multivariate design, or experimental design in principle properties as it is often referred to, is a combination of experimental design and PCA. Instead of variables, principle properties are used in design. This makes it possible to reduce the design considerably and still obtain relevant information, eg. which variables influence tablet quality. 3. RECENT OPTIMIZATION METHODS Factorial Design and Optimization Traditionally pharmaceutical formulations are developed by changing one variable at a time. The method is time consuming and it is difficult to evolve an ideal formulation using this classical technique since the combined effects of the independent variables are not considered (Baba et.al. 1989). It is therefore important to understand the complexity of pharmaceutical formulations by using established statistical tools such as factorial design. The number of experiments required for these studies is dependent on the number of independent variables selected (Dzido et.al. 1991). The response is measured for each trial and then either simple linear equation (4), or interactive equation (5) or quadratic (6) model is (Y = b0 + b1X1 + b2X2 + b3X3 ) (4)

Using PLS, it is possible to male predictions in Y and X data. The matrix is described by equation 2 and the Y matrix is described by equation 3. The inner relation then gives equation 4. SIMCA (Markowski W. 1993 & Wrisley L. 1993) SIMCA is a method used to classify objects. Provided the most of the variables express a real similarity, a class can be well approximated by a PC model with few components. From this model, which is the basis of the SIMCA method, a tolerance interval (usually 95%) can be constructed around the PC hyper planes, new objects are assigned to different classes depending on which tolerance cylinder they fit inside. All objects outside the tolerance interval are classified as outliers,

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

(Y = b0 + b1X1 + b2X2 + b3X3 + X1X2 + X1X3 + X2X3 +X1X2X3) (5) 2 2 (Y = b0 + b1X1 + b2X2 + b3X3 + b1 X11 + b2 X22 2 + b3 X33 + X1X2 + X2X3 + X1X3 + X1X2X3 ) (6) fitted by carrying out multiple regression analysis and F-statistic to identify statistically significant term. A prior knowledge and understanding of the process and the process variables under investigation are necessary for achieving a more realistic model (Schoenmakers et.al. 1991). Putting these equations an example it is tried to explain that with the help of these kinds of equations in factorial design methods the optimization can be achieved systemically (Jandera et.al. 1991). Global Optimization Global optimization is a branch of applied mathematics and numerical analysis that deals with the optimization of a function or a set of functions In real-life problems, functions of many variables have a large number of local minima and maxima (Adinarayana et.al. 2002). Finding an arbitrary local optimum is relatively straightforward by using local optimization methods. Finding the global maximum or minimum of a function is a lot more challenging and has been impossible for many problems so far (Akhnazarova et.al. 1982). Approaches (Allen et.al. 1992 & Anthony et.al. 1997) Deterministic The most successful are: Branch and bound methods Methods based on real algebraic geometry Stochastic, thermodynamics Several Monte-Carlo-based algorithms Simulated annealing Monte-Carlo with minimization Continuation Methods Related methods Tabu search Stochastic hill climbing Genetic algorithms Ant colony optimization The cross-entropy method

Worst case analysis Mathematical problems (e.g., the Kepler conjecture) The starting point of several molecular dynamics simulations consists of an initial optimization of the energy of the system to be simulated. Spin glasses The Simplex Optimization Methods The simplex methods are based on an initial design of k+1 trials, where k is the number of variables. A k+1 geometric figure in a k-dimensional space is called a simplex. The corners of this figure are called vertices (Spendley et.al. 1962). A simplex defined by three different trial conditions for two control variables. With two variables the first simplex design is based on three trials, for three variables it is four trials, etc (Nelder et.al. 1965). This number of trials is also the minimum for defining a direction of improvement. Therefore, it is a timesaving and economical way to start an optimization project (Walter et.al. 1991). After the initial trials the simplex process is sequential, with the addition and evaluation of one new trial at a time. The simplex searches systematically for the best levels of the control variables. The optimization process ends when the optimization objective is reached or when the responses cannot be improved further (Spendley et.al. 1962).

Applications of Global Optimization (Bolton et.al. 1997) Typical examples of global optimization applications include: Protein structure prediction (minimize the energy/free energy function) Traveling salesman problem and circuit design (minimize the path length) Chemical engineering (e.g., analyzing the Gibbs free energy) Safety verification, safety engineering (e.g., of mechanical structures, buildings)

1. 2. 3. 4.

Fig. 3. Control variable The Basic Simplex Method The Modified Simplex Method Evolutionary Operation Mixture Optimization

Artificial Neural Network An artificial neural network (ANN) or commonly just neural network (NN) is an interconnected group of artificial neurons that uses a mathematical model or computational model for information processing based on a connectionist approach to computation (Maass et.al. 1997). In most cases an ANN is an adaptive system that changes its structure based on external or

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

internal information that flows through the network. The term "neural network" can also mean biological-type systems. In more practical terms neural networks are non-linear statistical data modeling tools (Gurney et.al. 1997). They can be used to model complex relationships between inputs and outputs or to find patterns in data (Haykin et.al. 1999).

signal processing (Masters et.al. 1994). In some of these systems neural networks, or parts of neural networks are used as components in larger systems that combine both adaptive and nonadaptive elements (Smith, M. 1993). What they do however have in common is the principle of non-linear, distributed, parallel and local processing and adaptation. Employing artificial neural networks Perhaps the greatest advantage of ANNs is their ability to be used as an arbitrary function approximation mechanism which 'learns' from observed data (Wasserman, P. 1993). However, using them is not so straightforward and a relatively good understanding of the underlying theory is essential (Hertz et.al. 1990). Choice of model Learning algorithm. Robustness

Fig. 4. Artificial neural network A neural network is an interconnected group of nodes, akin to the vast network of neurons in the human brain. More complex neural networks are often used in Parallel Distributed Processing. Background of Artificial Neural Network There is no precise agreed definition among researchers as to what a neural network is, but most would agree that it involves a network of simple processing elements (neurons) which can exhibit complex global behavior, determined by the connections between the processing elements and element parameters (Hertz et.al. 1990). The original inspiration for the technique was from examination of the central nervous system and the neurons which constitute one of its most significant information processing elements (Lawrence, J. 1994). In a neural network model, simple nodes are connected together to form a network of nodes hence the term "neural network" (Masters et.al. 1994). These networks are also similar to the biological neural networks in the sense that functions are performed collectively and in parallel by the units (Smith, M. 1994). Currently, the term ANN tends to refer mostly to neural network models employed in statistics and artificial intelligence. Neural network models designed with emulation of the central nervous system (CNS) in mind are a subject of theoretical neuroscience (Wasserman et.al. 1993). In modern software implementations of artificial neural networks the approach inspired by biology has more or less been abandoned for a more practical approach based on statistics and

With the correct implementation ANNs can be used naturally in online learning and large dataset applications (Lawrence et. al. 1994). Their simple implementation and the existence of mostly local dependencies exhibited in the structure allows for fast, parallel implementations in hardware. Neural network software (Masters et.al. 1994 & Smith et.al. 1993) Neural network software is used to simulate, research, develop and apply artificial neural networks, biological neural networks and in some cases a wider array of adaptive systems. Simulators o Research simulators o Data analysis simulators Development Environments o Component based Criticism Custom neural networks Types of neural networks (Kesavan et.al. 1996, Wu T. et.al. 2000 & Chen et.al. 1999)

1. Feed forward neural network


Single-layer perceptron Multi-layer perceptron

2. Recurrent network Simple recurrent network Hopfield network 3. Stochastic neural networks 4. Modular neural networks Committee of machines 5. Holographic associative memory Instantaneously trained networks Cascading neural networks Neuro-fuzzy networks

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

A) Application of a Mixed Optimization Strategy in the Design of a Pharmaceutical Solid Formulation at Laboratory Scale The mixed strategy was used to optimize a dry powder blend containing 500 mg of alpha methyl dopa to be filled into hard gelatin capsules. The experimental plan consisted of assessing blend flow and dissolution rate using formulations manufactured at small laboratory scale, selecting the optimum formulation, and confirming the data. Two optimization techniques were used in the solid pharmaceutical product design: a genetic algorithm (GA) and a downhill simplex technique. B) Multivariate spline interpolation as a novel method to optimize pharmaceutical formulations Fig.5. A two-layer neural network capable of calculating XOR. Applications The utility of artificial neural network models lies in the fact that they can be used to infer a function from observations. This is particularly useful in applications where the complexity of the data or task makes the design of such a function by hand impractical (Khuri et. al. 1981). Real life applications (Barron A.R. 1993) The tasks to which artificial neural networks are applied tend to fall within the following broad categories: Function approximation, or regression analysis, including time series prediction and modelling. Classification, including pattern and sequence recognition, novelty detection and sequential decision making. Data processing, including filtering, clustering, blind source separation and compression. The generation of response surfaces using multivariate spline interpolation (MSI) has provided rapid and detailed information. Nevertheless, no application of MSI in the pharmaceutical field has been reported to date, even though it promises potential applications. To overcome the shortcomings of the classical response surface method, newly a multi-objective simultaneous optimization method was developed, in which MSI had been incorporated. The method was applied to the optimization problem of a transdermal hydrogel formulation for ketoprofen containing several chemical enhancers. Results suggested a superior function of the MSI approach. C) Formulation Optimization of Paclitaxel Carried by PEGylated Emulsions Based on Artificial Neural Network (Patterson D. 1996 & Baba Y et. al. 1999) An artificial neural network (ANN) was used to optimize the formulation of paclitaxel carried by injectable PEGylated emulsions; which has a small particle size, high entrapment efficiency and good stability. A computer optimization technique based on a spherical experimental design for three-level, three factors [soybean oil (X1), PEG-DSPE (X2) and polysorbate 80 (X3)] were used to optimize the formulation. Novel formulation for paclitaxel emulsions was optimized with ANN and prepared. The contribution indices of each component suggested that PEGDSPE mainly contributes to the entrapment efficiency and particle size of paclitaxel emulsions, while polysorbate 80 contributes to stability. 3. FUTURE PROSPECTIVES The scope of optimization technique is intended to support innovation and efficiency in pharmaceutical industry. The framework is founded on understanding to facilitate innovation and production. Pharmaceutical companies need to adopt new technologies, processes and collaborations (Baba Y et. al. 1989).

Application areas (Bishop CM. 1995) system identification and control game-playing and decision making pattern recognition sequence recognition medical diagnosis financial applications data mining visualization and e-mail spam filtering

Examples The fallowing example can illustrate importance of optimization techniques in pharmaceutical formulation

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

Software and data analysis tools Networks of research alliances to ensure access to new genetic technologies and research Research scientists working in collaboration as virtual drug discovery teams Standardized tools and processes delivered through drug discovery portals e.g. gene mapping.

4. CONCLUSION The field of optimization is broad and has applications in all areas of pharmaceutical science. Naturally, many different methods have been used and in some cases even developed by chemists. In this article, an overview is given where the methods have been classified as conventional and recent methods. The former category comprises methods that are very good in dealing with low-dimensional problems. An example is the optimization of a chromatographic experiment, where optimal separation conditions should be found in as few experiments as possible. These methods are said to be strong because they incorporate much knowledge from the user about the problem at hand. The conventional category also includes weak methods where knowledge is not a prerequisite. As the efficiency of these methods is in many cases very low, they are not often used in industry. The recent optimization techniques described in this article have been specifically developed for those cases where the conventional techniques were not suitable. Because the number of evaluations may be quite high they usually are applied in connection with computer experiments rather than with laboratory experiments. The conventional and recent methods have complementary roles in pharmaceutical industry. With the increasing role of computational techniques, the recent methods will continue to flourish. Many standard computer software packages are available for experimental design, and are sometimes even included in laboratory instruments. Although the user should still understand the principles of the methods, no programming is required, thereby significantly lowering the threshold for their use. 5. REFERENCE
Adinarayana, K., Ellaiah, P. Response surface optimization of the critical medium components for the production of alkaline protease by a newly isolated bacillus sp. J. Pharm Pharmaceut Sci. 2002; 5(3): 281287. Akhnazarova, S., Kafarov, V. Experiment optimization in chemistry and chemical engineering, Mir Publications, Moscow, 1982. Allen, T.M., Mehra, T., Hansen, C., Chin, Y.C. Stealth Liposomes: an improved sustained release system for 1--Darabinofuranosylcytosine. Cancer Res. 1992, 52, 2431. Anthony A. N. and James, K.C. Pharmaceutical experimental design and interpretation, Taylor and Francis Publishers, Bristol PA USA, 1996, 131-192.

Baba Y and Ito M K. Optimization of gradients in anionexchange separations of oligo nucleotides using computer-assisted retention prediction and a high-performance liquid chromatographic simulation system. J Chromatogr A. 1989, 485: 647-655. Baba Y and Kura G. Computer-assisted retention prediction system for inorganic cyclic polyphosphates and its application tooptimization of gradients in anion-exchange chromatography. J Chromatogr A,550: 514. 1991. Barron, AR. Universal approximation bounds forsuper position of a sigmoidal function. IEEE Transaction on Information Theory,1993, 39:90-945. Bishop C.M. Neural networks for pattern recognition. Oxford University Press, Oxford, London, 1995. Bolton, S. Pharmaceutical statistics: practical and clinical applications. 3rd edn, Marcel Dekker Inc., New York, 1997; 217-241. Chen Y., McCall T.W., Baichwal A.R., Meyer MC. The application of an artificial neural network and pharmacokinetic simulations in the design of controlled-release dosage forms. J Control Release. 1999; 59:33-41. Chloupek R.C., Hancock W.S., Snyder L.R. Computer simulation as a tool for the rapid optimization of the high-performance liquid chromatographic separation of a tryptic digest human growth hormone. J Chromatogr A. 1992; 594: 65-73. Cornell, J. A. Experiments with Mixtures. John Wiley & Sons, New York, USA, 1990. De Boer J.H. Smilde A.K., Doornbos D.A., Introduction of multi-criteria decision making in optimization procedures for pharmaceutical formulations, Acta Pharrn. Technol. 1988; 34-37. De Boer J.H., Bolhuis G.K., Doornbos D.A., Comparative evaluation of multicriteria decision making and combined contour plots in optimization of directly compressed tablets, Eur. J. Pharm. Biopharm. 1991; 37. Dick C.F., Klassen R.A., Amidon G.E., Determination of the sensitivity of a tablet formulation to variations in excipient levels and processing conditions using optimization techniques, Int. J. Pharm. 1987; 38-40. Doornbos D.A., Optimization in pharmaceutical sciences, Pharm. Weekblad Scientific Ed.3rd, 1981. Dzido T. H., Soczewinski E., Gudej J. Computer-aided optimization of high-performance liquid chromatographic analysis of flavonoids from some species of the genus. J Chromatogr A, 1991; 550: 71-76. Gurney, K. An Introduction to neural networks, Routledge, London,1997. Haykin, S. Neural networks: A comprehensive foundation, Prentice Hall, 1999. Hertz, J., Palmer, R.G., Krogh. A.S. Introduction to the theory of neural computation, Perseus Books, 1990. Hobbs, S.K., Monsky, W.L., Yuan, F., Roberts, W.G., Griffith, L., Torchilin, V.P. Jain, R.K. Regulation of transport pathways in tumor vessels: Role of tumor type and micro environment. Proc. Natl. Acad. Sci. USA. 1998; 95: 4607-461. Jandera P. Predictive calculation methods for optimization of gradient elution using binary and ternary solvent gradients. J Chromatogr A. 1989; 485: 113-141. Kesavan J.G., Peck G.E. Pharmaceutical granulation and tablet formulation usingneural networks. Pharm. Dev. Technol. 1996; 1:391-404.

Journal of Current Pharmaceutical Research 2011; 7 (1): 21-28

Khuri I., Conlon M. Simultaneous optimization of multiple responses represented by polynomial regression functions. Technometrics. 1981; 23: 363-375. Lawrence, J. Introduction to neural networks, California Scientific Software Press. 1994. Liu C.L., Zhu P.L., Liu M.C. Computer-aided development of a high-performance liquid chromatographic method for the determination of hydroxyanthraquinone derivatives in Chinese herb medicine rhubarb. J. Chromatogr A, 1999; 857: 167-174. Maass, W. Fast Sigmoidal Networks via spiking neurons. neural computation 1997; 9: 279-281. Mackay D.J. Information theory, inference and learning algorithms. Prepublication, 1998. Markowski W. Computer-aided optimization of gradient multiple development thin-layer chromatography (III): Multi-stage development over a constant distance. J Chromatogr A, 1996; 726: 185-192. Markowski W. Computer-aided optimization of gradient multiple development thin-layer chromatography (II): Multi-stage development. J. Chromatogr. A., 1993; 635: 283-289. Masters T. Signal and image processing with neural networks, John Wiley & Sons, Inc., 1994. Nelder J.A., R. Mead, A simplex method for function optimization, Computer J., 1965; 7, 308-313. Papahadjopoulos D. Gabizon A.A. Sterically stabilized liposomes: pharmacological properties and drug carrying potential in cancer, in liposomes as tools in basic research and industry, CRC Press, Boca Raton, FL. 1995; 177-188. Patterson D. Artificial neural networks. Singapore: Prentice Hall; 1996. Press W.H., Flannery B.P., Teukolsky S.A., Vetterling W.T., Numerical recipes in C, Cambridge University Press, Cambridge, 1988.

Schmidt A.H., Molnar I. Computer-assisted optimization in the development of a high-performance liquid chromatographic method for the analysis of kava pyrones in Piper methysticum preparations. J. Chromatogr A, 2002; 948: 51-63. Schoenmakers P. J., Bartha A., Billiet H.A.H. Gradient elution methods for predicting isocratic conditions. J Chromatogr A, 1991; 550: 425-447. Smith, M. Neural networks for statistical modeling, Van Nostrand, Reinhold, 1993. Spendley, G.R., Hext, F.R., Himsworth, Sequential application of simplex designs in optimization and evolutionary operations, Technometrics, 1962; 4, 441-461. Takayama K., Fujikawa M., Nagai T. Artificial neural networks as a novel method to optimize pharmaceutical formulations. Pharm Res. 1999;1-6. Tricot, G.. De Bock, R., Dekker, A.W. Low dose cytosine arabinoside in myelodysplastic syndromes. Br. J. Haematol. 1984; 58, 231-240. Walter F.H., Parker L.R., Morgan S.L., Deming S.N. Sequential simplex optimization, CRC Press, Boca Raton, 1991. Wasserman, P. Advanced methods in neural computing, Van Nostrand Reinhold, 1993. Wolcott R.G., Dolan J. W., Snyder L. R. Computer simulation for the convenient optimization of isocratic reversed phase liquid chromatographic separations by varying temperature and mobile phase strength. J Chromatogr A, 2000; 869: 3-25. Wrisley L. Use of computer simulations in the development of gradient and isocratic high-performance liquid chromatography methods for analysis of drug compounds and synthetic intermediates. J Chromatogr A, 628: 19119,1993. Wu T., Pan W., Chen J., Zhang R. Formulation optimization technique based on artificial neural network in salbutamol sulfate osmotic pump tablets. Drug. Dev. Ind. Pharm. 2000; 26:211-215.

You might also like