Computer-Aided Formulation Development (CADD): Optimization, DOE, and Pharma Applications

Modern drug formulation requires optimization, efficiency, and reproducibility. This deck introduces CADD techniques and statistical tools such as factorial design, response surface methodology, Taguchi design, Plackett–Burman, and Latin square design. Applications in emulsions, microemulsions, and carrier systems are illustrated with real case studies, highlighting how in-silico and experimental design approaches can reduce costs, improve product performance, and enhance innovation. It is particularly valuable for students working on formulation projects and research dissertations.
This slide deck introduces Computer-Aided Formulation Development (CADD), focusing on optimization techniques, factorial design, response surface methodology, and experimental screening designs. It explains independent vs. dependent variables, constrained vs. unconstrained problems, and statistical designs such as Plackett-Burman, Taguchi, Latin Square, Central Composite, and Box-Behnken designs. Applications in pharmaceutical emulsions, microemulsions, and carrier systems are illustrated with case studies. Students will understand the importance of in-silico tools, DOE (Design of Experiments), and optimization methods to reduce costs, save time, enhance reproducibility, and ensure quality. This resource is useful for M.Pharm students, researchers, and formulation scientists who want to strengthen their foundation in pharmaceutical statistics and computational modeling.

1. Computer-Aided Formulation Development Presented by Ayushi Jain M.Pharm 1

2. Concept of optimization Optimization parameters Factorial design Optimization technology Screening design Computers in pharmaceutical formulation Development of pharmaceutical emulsion Development of pharmaceutical Microemulsion carriers 2

3. Contents Optimization Optimization necessary Terms Optimization parameters Types of experimental designs Factorial design Response surface design forms of optimization techniques Application of Computer – Aided Techniques in Development of a)Pharmaceutical Emulsions b)w/o Emulsions c) MicroEmulsion Drug Carriers Example 3

4. Optimization It is defined as to make perfect, effective, or functional as possible. It is process of finding the best way of using the existing resources while taking into the account of all factors that influences decisions in any experiment. Traditionally, optimization in pharmaceuticals refers to changing one variable at a time so as to obtain solution of a problematic formulation Modern pharmaceutical optimization involves systematic design of experiments (DOE) to improve formulation irregularities. In the other word we can say that – quantitate a formulation that has been qualitatively determined. Its not a screening technique Optimization makes the perfect formulation & reduce the cost Primary objective may not be optimize absolutely but to compromise effectively & there by produce the best formulation under a given set of restrictions 4

5. Why Optimization Is Necessary? Optimization Reduce the cost Save the time Safety & reduce the error Innovation & efficacy Reproducibility 5

6. Terms used Factor : it is an assigned variable such as concentration, temperature etc Quantitative : numerical factor assigned to it Ex :- concentration – 1%, 2%, 3% etc… Qualitative : which are not numerical Ex :- polymer grade, humidity condition etc. Levels : levels of a factor are the values or designations assigned to factor Factor levels Temperature 30 Concentration 1%, 2% Factor levels Temperature Concentration 1%, 2% 6

7. response : it is an outcome of the experiment It is the effect to evaluate Ex :- disintegration time etc. Effect : it is the change in response caused by varying the levels It gives the relationship between various factors & levels Interaction : it gives the overall effect of two or more variables Ex :- combined effect of lubricant & glidant on hardness of the tablet 7

8. optimization parameters Problem type Variable Constrained Unconstrained Independent Dependent Formulating variable Processing variable 8

9. Problem type a) Constraints Ex :- making hardest tablet but should disintegrate within 20 minutes b) Unconstraint Ex :- making hardest tablet Variables type a) Independent variable Ex :- mixing time for given process step granulating step b) Dependent variable Which are the responses or characteristics of the in process material Ex :- particle size of the vesicles, hardness of tablets 9

10. Higher the number of variables, more complicated will be the optimization process. There should be a relationship between the given response & the independent variable, & once this relationship is established, a response surface is generated. From response surface only, we find the points which will give desirable value of the response. Example of dependent & independent variables 10 Independent variable Dependent variable Compressional force Mixing rate Mixing time Disintegration time Hardness Dissolution Friability Weight uniformity

11. 11

12. Factorial Design These are the designs of choice for simultaneous determination of the effects of several factors & their interactions. Symbols to denote levels are: when both the variables are in low concentration. a - one low variable and second high variable. b - one high variable and second low variable. ab - both variables are high. Factorial designs are optimal to determined the effect of pressure & lubricant on the hardness of a tablet. 12

13. Effect of disintegrate & lubricant concentration on tablet dissolution. It is based on theory of probability and test of significance. It identifies the chance variation ( present in the process due to accident) and the assignable variations ( which are due to specific cause ). Factorial design are helpful to deduce IVIVC. IVIVC are helpful to serve a surrogate measure of rate and extent of oral absorption. BCS classification is based on solubility and permeability issue of drugs, which are predictive of IVIVC. Sound IVIVC omits the need of bioequivalence study. IVIVC is predicted at three levels : 13

14. Level A - point to point relationship of in vitro dissolution and in vivo performance. Level B - mean in vitro and mean in vivo dissolution is compared and co-related. Level C - correlation between amount of drug dissolved at one time and one pharmacokinetic parameter is deduced. BCS Classification and its expected outcome on IVIVC for Immediate release formulation BCS Class Solubility Permeability IVIVC I High High Correlation( if dissolution is rate limiting ) II Low High IVIVC is expected III High Low Little or no IVIVC IV low Low Little or no IVIVC 14

15. Types of Factorial Design Full Used for small set of factors b) Fractional It is used to examine multiple factors efficiently with fewer runs than corresponding full factorial design Types of fractional factorial designs Homogenous fractional Mixed level fractional Box-Hunter Plackett – Burman Taguchi Latin square 15

16. Homogenous fractional Useful when large number of factors must be screened Mixed level fractional Useful when variety of factors needed to be evaluated for main effects and higher level interactions can be assumed to be negligible. Ex :- objective is to generate a design for one variable A at 2 levels and another X at three levels, mixed & evaluated. Box-Hunter Fractional designs with factors of more than two levels can be specified as homogenous fractional or mixed level fractional 16

17. Plackett-Burman It is a popular class of screening design. These designs are very efficient screening designs when only the main effects are of interest. These are useful for detecting large main effects economically ,assuming all interactions are negligible when compared with important main effects Used to investigate n-1 variables in n experiments proposing experimental designs for more than seven factors. Taguchi It is similar to PBDs. It allows estimation of main effects while minimizing variance. Taguchi Method treats optimization problems in two categories, 17

18. Static Problems :- Generally, a process to be optimized has several control factors which directly decide the target or desired value of the output. Dynamic Problems :- If the product to be optimized has a signal input that directly decides the output. Latin square They are special case of fractional factorial design where there is one treatment factor of interest and two or more blocking factors. Signal-to-Noise ratios (S/N), which are log functions of desired output. 18

19. We can use the Latin square to allocate treatments. If the rows of the square represent patients and the columns are weeks, then for example the second patient, in the week of the trial, will be given drug D. Now each patient receives all five drugs, and in each week all five drugs are tested. 19

20. Response Surface Designs This model has quadratic form γ =β0 + β1X1 + β2X2 +….β11X12 + β22X22 Designs for fitting these types of models are known as response surface designs. If defects and yield are the outputs and the goal is to minimize defects and maximize yield. Two most common designs generally used in this response surface modeling are a) Central composite designs b) Box- Behnken designs 20

21. a) Box-Wilson central composite Design This type contains an embedded factorial or fractional factorial design with centre points that is augmented with the group of ‘star points’. These always contains twice as many star points as there are factors in the design. The star points represent new extreme value (low & high) for each factor in the design. To picture central composite design, it must imagined that there are several factors that can vary between low and high values. 21

22. Central composite designs are of three types Circumscribed(CCC) designs-Cube points at the corners of the unit cube ,star points along the axes at or outside the cube and centre point at origin. Inscribed (CCI) designs-Star points take the value of +1 & -1 and cube points lie in the interior of the cube. Faced(CCI) –star points on the faces of the cube. 22

23. b) Box- Behnken design Box- Behnken designs use just three levels of each factor. In this design the treatment combinations are at the midpoints of edges of the process space and at the center. These designs are rotatable (or near rotatable) and require 3 levels of each factor. These designs for three factors with circled point appearing at the origin and possibly repeated for several runs. It’s alternative to CCD. The design should be sufficient to fit a quadratic model, that justify equations based on square term & products of factors. Y= b0+b1x1+b2x2+b3x3+b4x1x2+b5x1x3+b6X2X3+b7X12+b8X22+b9X32 23

24. Three-level full factorial designs It is written as 3k factorial design. It means that k factors are considered each at 3 levels. These are usually referred to as low, intermediate & high values. These values are usually expressed as 0, 1 & 2. The third level for a continuous factor facilitates investigation of a quadratic relationship between the response and each of the factors. Three level full factorial design for 3 factors, 3 designs with 27 experiments. 24

25. Forms of Optimization techniques 1. Sequential optimization techniques. 2. Simultaneous optimization techniques. 3. Combination of both. Sequential Methods Also referred as the "Hill climbing method". Initially small number of experiments are done, then research is done using the increase or decrease of response. Thus, maximum or minimum will be reached i.e. an optimum solution. 25

26. Simultaneous Methods Involves the use of full range of experiments by an experimental design. Results are then used to fit in the mathematical model. Maximum or minimum response will then be found through this fitted model. Ex :- Designing controlled drug delivery system for prolonged retention in stomach required optimization of variables like presence/ absence / concentration of stomach enzyme, pH, fluid volume and contents of guts, Gastric motility and gastric emptying. When given as single oral tablet (A). Same drug when given in multiple doses (B). Same drug when given as optimized controlled release formulation (C). 26

27. Application of Computer – Aided Techniques in Development of Pharmaceutical Emulsions Emulsions are disperse systems made of two immiscible liquids. One liquid is dispersed into the other, in the presence of surface active agents such as emulsifiers. The two liquids are usually oil & water. Simple emulsions are oil-in-water(o/w) or water-in-oil (w/o). In pharmaceutical industry it is been have great potential – topical, parenteral or oral. However emulsions are thermodynamically unstable systems & different phenomena during storage could occur, including gravitational separation ( creaming/sedimentation ), flocculation, coalescence, Ostwald ripening & phase inversion 27

28. Stability & properties of emulsions are influenced by different factors, formulation & process optimization techniques would be useful for finding the ideal emulsion formulation. The main parameters relating to the stability, effectiveness & safety of the pharmaceutical emulsion should be optimized simultaneously More intense application of in- silico techniques were applied Prinderre applied factorial design methods to optimize the stability & suggested the required hydrophilic lipophilic balance (HLB) of (o/w) emulsions prepared with sodium lauryl sulphate as the surfactant 28

29. Experimental design determined the required HLB with good approximation in five runs for the average diameter & viscosity studies, while the conductivity study needed at least eight runs 29

30. A – mixing rate B – homogenisation C – mixing time   30 A B C - - - - A + - - B - + - C - - + AB + + - AC + - + BC - + + ABC + + +

31. Simovic investigated the influence of the processing variables on performance of o/w emulsion gels stabilized by a polymeric emulsifier. A two-factor three-level experimental design at two sets was applied :- using a laboratory mixer & a dispenser. Independent variables were mixing speed & time. Dependent variables were millimetres of oil phase separated after centrifugation at 3500rpm in a laboratory centrifuge, & viscosity at shear rate of 180l/s. The responses were fitted into a second order model by means of multiple regression analysis. It was defined that the most favourable condition for preparing stable o/w emulsions using the laboratory mixer with a mixing speed at 1500rpm & mixing time of 20 minutes. 31

32. Two factor three level experimental design By response surface methodology Two factors :- A & B Three levels :- -1, 0 & +1   32 A B -1 -1 +1 -1 +1 +1 -1 +1

33. Rahali found the optimal preservative combination & concentration for preparing topical emulsion by using a D-optimal experimental design (mixture design) . In this study three preservatives were tested benzoic acid, sorbic acid, & benzyl alcohol. The preservative effects were evaluated using the antimicrobial preservative efficacy test(challenge test ) of the EP. The results of this study were analysed with the help of design expert software. From this it was able to formulate topical emulsions in accordance with the requirements of the EP. Study of the influence of different factors for emulsion system is difficult, due to the problems of complicated nonlinear correlations. The Artificial Neural Networks (ANN) technique used to solve this problems. 33

34. Gasperlin investigated the influence of different ratios of individual components on the viscoelastic behaviour of semisolid lipophilic emulsion systems using this technique. The creams were prepared according to preliminary experimental designs ( mixture designs ). ANN involved 3 input, 12 hidden & 9 output neurons. Input neuron – silicone surfactant, purified water, white petroleum. Output neurons – rheological parameter tan . Similarly ANN used for prediction of the complex dynamic viscosity of semi-solid w/o emulsions.   34

35. Also it is used to optimize ( kumar ) the fatty alcohol concentration in the formulation of an o/w emulsion. Emulsion was composed of purified water, liquid paraffin, sodium lauryl sulphate & fatty lauryl alcohol. Predictions from ANN,s are accurate & allow quantification of the relative importance of the inputs. Further by varying the network topology & parameters it was possible to obtain output values (zeta potential, viscosity, conductivity, & particle size ) that were close to experimental values. The ANN model predicted results & the actual output values were compared. An R2 value of 0.84 for the ,model suggested adequate modelling which is supported by the correlation coefficient of 0.9445. 35

36. 36 ANN

37. Orthogonal experimental design for w/o emulsion optimal formulation composition of insulin emulsion predicted by MSI (multivariate spline interpolation) & the study was performed using the model formulation prepared according to the two-factor spherical second order composite experimental design. The volume ratio of the outer aqueous phase and the agitation time were selected as the causal factors. 37

38. Multiple emulsion w/o/w Oil globules containing small droplets of water is dispersed in an aqueous continuous phase. The stability is affected by a number of factors including the method of preparation, osmotic balance between the internal & external water phase, phase volume ratio, type & concentration of the emulsifiers. Onuki in his work optimize the formulation of w/o/w multiple emulsion incorporating insulin was performed based on statistical method such as the orthogonal experimental design & the response surface method. 16 types of emulsions were prepared according to the orthogonal experimental design. Input – amount of gelatin, insulin, oleic acid, volume ratio of the outer water phase & agitation time of the second emulsification process. 38

39. Output – inner droplet size, viscosity, stability & pharmacological effect. Based on analysis of variance (ANOVA) it as concluded that the most predominant contribution of all causal factors was the volume ratio of the outer water phase. Used simultaneous optimization technique, in which multivariate spline interpolation (MSI) was incorporated. A two - factor composite second order spherical experimental design was used to select model formulation analysed by dataNESIA (Tokyo ). It was predicted that the pharmacological activity & stability as high as a pharmaceutical formulation. Other experimental designs techniques for w/o/w emulsions are Statisitca Back propagation (BP) 39

40. Application of Computer - Aided Techniques in Development of Microemulsion Drug Carriers Microemulsions are thermodynamically stable & optically isotropic transparent colloidal system consist of oil, water & surfactant. They may be o/w, w/o or bicontinuous. Microemulsion & self- microemulsifying drug delivery systems (SMEDDS) form only in well balanced mixtures of the selected excipients & within specific concentration ranges of the constituents at given temperature & pressures ( i.e. the microemulsion area ). Due to phase behaviour differentiation of the multicomponents ( surfactant, cosurfactant , oil, water & drugs ) requires large number of experiments. Further characterization of a microstructure is a difficult task, due to its dynamic character as well as nanoscale organization. 40

41. ANN model was used for differentiation & prediction of the microemulsion forming systems. Richardson demonstrated the use of ANNs to identify the physicochemical properties of the cosurfactant with relevance for microemulsion formation in the four-compartment system lecithin (surfactant), isopropyl myristate (oil), triple distilled water & cosurfactant. The cosurfactant (i.e. short & medium chain alcohols, amines, acids & ethylene glycol monoalkyl ethers ) were employed. The BP feed forward algorithm of learning & four computed surfactant molecule properties ( molecular volume, areas for its head group, hydrophobe & computed octanol /water logP value ) were selected. The output was presence (+1) or absence (-1) of microemulsion formation in a particular mixture. 41

42. ANN was extracted from the pseudo-ternary diagrams. Software was YANNI . Diagram consists of surfactant to cosurfactant which involves 6 input neurons, a single layer of hidden 14 neurons & 1 output. The trained MLP ( ANNs simulator software, NNMODEL version 1.404. Neural Fusion ) predicted phase behavior of quaternary systems. Where MLP is multilayer perceptron MLP architecture with a BP learning rule ( Neural Networks, StatSoft Inc , Tulsa USA ). Genetic algorithm (GA) ( Pallas Compu Drug Int. San Francisco, USA ). ChemSketch ( ACD Inc , Toronto Canada ). Neuro-Fuzzy Modelling (NFM) by using neural networks simulator Nets2014. Radial Basis Function (RBF) 42

43. Neuro Fuzzy 43

44. EXAMPLE Formulation and optimization of microemulsion -based organogels containing propranolol hydrochloride using experimental design methods Response Surface Methodology (RSM) is a rapid technique used to derive a functional relationship between an experimental response and a set of input variables empirically. By using RSM, the number of experimental runs that is necessary for the establishment of a mathematical trend in the experimental design region will be reduced, allowing to determine the optimum level of experimental factors required for a given response. 44

45. Construction of partial pseudo-ternary phase diagram :- Samples containing different weight ratios of lecithin/IPM (20:80, 30:70, 40:60, 50:50 and 60:40) were initially prepared. Phase studies were carried out by adding either pure water or solutions of PR with various concentrations (10, 20, 30, 50% w/v) to the mixture of lecithin/IPM, while stirring. After addition of each 10 µl aqueous solution, resulting systems were examined for clarity and viscosity. Modified simplex :- Modified Simplex was used for the sequential design of experiments and optimization. In this study, three factors (k), including lecithin, PR and water were selected. IPM, as another factor, was considered as 100- % lecithin. In the first step, we started with four (k+1) runs and moved forward to the desired responses through measuring various projections of the rejected trial condition (W), Reflection (R), Expansion (E), CR or C+( positive contraction) and CW or C- (negative contraction). The narrower region of these factors corresponding to near optimum conditions was further used to construct RSM design. 45

46. RSM (Central Composite) design :- The experimental data was analyzed by response surface regression procedure and the results were statistically analyzed by the corresponding analysis of variances of the selected experimental design. If three factors are studied at two levels, the relevant equation would be equation 1, with three two-way Independent variables and their constraints. Independent variables Unit Lower constraint Upper constraint lecithin* % 30 50 PR % 20 40 water % 3 4 interactions and one three-way interaction. Y=β0+β1X1+β2X2+β3X3+β12X1X2+β13X1X3+β23X2X3+β123X1X2X3+ ε If a quadratic relationship is sought, would be used. Y=β0+β1X1+β11X12+β2X2+β22X22+β3X3+β33X32+β12X1X2 + Β13 x1x3 +β23X2X3+β123X1X2X3+ε 46

47. Partial phase diagram of organogels containing lecithin/IPM and water at lecithin/IPM weight ratio of 60:40. 47

48. Factor 1 Factor 2 Factor 3 Response 1 Response 2 Std Run Point Type Block A:lecithin B:PR C:Water Flux Viscosity % % % microgram/cm2/ hr p 1 7 Fact Block 1 50 40 3 360 560 2 4 Fact Block 1 50 30 4 350 510 3 1 Fact Block 1 30 40 4 500 460 4 5 Fact Block 1 30 30 3 505 366 5 3 Center Block 1 40 35 3.5 469 402 6 6 Center Block 1 40 35 3.5 490 377 7 2 Center Block 1 40 35 3.5 470 385 8 8 Axial Block 2 25 35 3.5 479 375 9 16 Axial Block 2 54 35 3.5 360 620 10 13 Axial Block 2 40 28 3.5 440 379 11 11 Axial Block 2 40 42 3.5 380 419 12 9 Axial Block 2 40 35 2.8 487 380 13 10 Axial Block 2 40 35 4.2 400 397 14 15 Center Block 2 40 35 3.5 483 376 15 14 Center Block 2 40 35 3.5 478 395 16 12 Center Block 2 40 35 3.5 486 370 48

49. Plots of viscosity versus X1: lecithin, X2: PR at 3.5% water as the actual factor; a) contour plot; b) 3D plot. 49

50. 3D plots of the predicted formulations (X1: Lecithin, X2: Pr ) at a) 3.2% water content (C: Water = 3.20) ; b) 3.1% water content (C: Water = 3.10). 50

51. Overlay plots of responses (flux & viscosity) for predicted formulations at various water content as the actual factor (X1: Lecithin, X2: Pr ); a) 3.1% (C: Water = 3.00); b) 3.2% (C: Water = 3.20) and c) 3.5% (C: Water = 3.50). The light gray region stands for formulations with maximum flux and minimum viscosity. 51

52. Experimentally prepared formulations based on the predicted results and the evaluation of flux and viscosity.   Components   Flux (μg/cm 2 /h) Viscosity (p)   Number Lecithin (%) PR (%) Water (%) Predicted Observed Predi cted Observed Euclidean distance 1 34.69 31.76 3.03 514 510 354 350 5.6 2 25.2 30.77 3.15 507 503 358 361 5 3 36.35 32.53 3.21 507 505 356 358 2.8 4 35.16 31.64 3.05 513 510 356 348 8.5 5 36.94 33.12 3.1 510 508 360 353 7.3 52

53. CONCLUSION :- The main objective of the present study was to prepare a lecithin-based microemulasion gel ( organogel ) containing propranolol HCl with a predictable release rate. Data confirms that the choice of lecithin/IPM weight ratio and the amount of drug incorporated may be crucial in determining the performance of a microemulaion -based gel. Experimental design methods were used to optimize the release of PR from the gel. The tested parameters were lecithin, PR and water contents and the gel viscosity and flux were considered as responses for optimization. Our study demonstrates that experimental design technique is a valuable tool for optimization of organogel formulations, which enables to have a better understanding of how different, crucial variables could influence the selected responses. 53

54. THANK YOU 54

55. References 1. Computer Applications in Pharmaceutical Research and Development, Sean Ekins , 2006, John Wiley & Sons. 2. Computer-Aided Applications in Pharmaceutical Technology, 1 ST Edition, Jelena Djuris , Woodhead Publishing. 55