Publication: Yapay sinir ağlarıyla difenhidramin HCL içeren matriks tabletlerin modellenmesi ve optimizasyonu
Abstract
YAPAY SİNİR AĞLARIYLA DİFENHİDRAMİN HCL İÇEREN MATRİKS TABLETLERİN MODELLENMESİ VE OPTİMİZASYONU ÖZET Bu çalışmada, etkin madde olarak bir etanolamin türevi olan difenhidramin HCl ve yardımcı madde olarak hidroksipropil metil selüloz(HPMC)’un iki farklı viskoziteye sahip polimeri, mikrokristalin selüloz (Avicel) ve laktoz kullanılmıştır. Hidrolik el presi kullanılarak direk basım yöntemiyle matriks tabletler elde edilmiştir. Formüllerde etkin madde miktarı 50 mg, tablet basım kuvveti 850 kg-f olarak sabit tutulup, 4 adet yardımcı madde değişik miktarlarda toplamda 325 mg kullanılmıştır. Yardımcı maddelerin farklı miktarlarda kullanılışı tablet kırılma kuvveti, etkin maddenin %50’sinin salımı için geçen süreyi ve 5. saatte salınan % ilaç miktarını değiştirmiştir. Yardımcı madde olarak HPMC100K, HPMC4K, Avicel ve laktoz, bağımsız değişken olarak seçilmiş ve bu bağımsız değişkenler farklı formüllerin içinde farklı miktarlarda kullanılmıştır. Bağımlı değişken olarak seçilen tablet kırılma kuvvetini etkin maddenin %50’sinin salımı için geçen süre ve 5. saatte salınan % ilaç miktarının değişimi izlenmiştir. Bağımlı değişkenlerin değişmesine neden olabilecek bağımsız değişkenlerin dışındaki etmenlerin etkisini en aza indirmek için karıştırma süresi ve tablet sıkıştırma kuvveti her formül için sabit tutulmuştur. 32 formül için kontrol edilen bağımlı değişkenler Yapay Sinir Ağları (YSA) modellenmiştir. Modelleme sırasında bağımlı değişkenlerimizden 2’li, 3’lü ve 4’lü setler oluşturarak en iyi yaklaşıklığı sağlayan regresyon sayısına ulaşılmaya çalışılmıştır. Bu amaçla bağımlı değişkenleri en iyi şekilde tahmin edebilen 2 model regresyon katsayısı 1’e en yakın olanlar içinden seçilmiştir. Seçilen modellerle tekrar yapay sinir ağları stediğimiz kriterleri sağlayan tabletin optimizasyonunu gerçekleştirilmiştir. Seçilen modellerle oluşturulan optimize tabletlerin istenilen kriterleri; 9 kg-f kırılma kuvvetini sağlaması, etkin maddenin 2. saatte %50’sinin, 5. saatte %70 inin salımını gerçekleştirebilmesidir. Modellerle optimize edilen iki ayrı formül kullanılarak tabletler test edilmiştir. Kontrol deneylerinde tablet kırılma kuvvetinde %100 yakın, etkin maddenin %50’sinin salımı için geçen sürede: %91 ve 5. saatte salınan % ilaç miktarında ise %96,4 başarı sağlanmıştır. Anahtar kelime: Yapay sinir ağları (YSA), Modelleme, Optimizasyon, Difenhidramin, Tablet Modelling and Optimization of Matrix tablets containing diphenhidramine HCl by using Artificial Neural Network
Modelling and Optimization of Matrix tablets containing diphenhidramine HCl by using Artificial Neural Network In this study, diphenhidramine HCl as the active substance, an etanolamine derivative, and hydroxypropylmethyl cellulose (HPMC) polymer with two different viscosity, microcrystalline cellulose (Avicel) , lactose as excipients were used. The matrix tablets were prepared by direct compression method with hydrolic hand pressing equipment. Thirty two different tablet formulations were designed based on one master formula. The amount of active substence, diphenhidramine HCl of 50 mg and compression force of tablet 850kg-f were kept constant while the excipients were used in different amounts in sum of 325 mg. Different mixtures of excipients have changed tablet crushing strength, period to release 50% of the active substance and released active substence amount in fifth hour. In experiments, excipients HPMC100K, HPMC4K, Avicel and lactose were chosen independent variable while tablet crushing strength, period to release 50% of the active substance and released active substence amount in fifth hour were chosen as dependent variables. Then changes in dependent variables were observed. In order to maintain minumim side effects on dependent variables, mixing period and pressing force were kept constant on formulations. Controlled dependent variables in thirty two formulas were applied to the artificial neural network program for modelling. In modelling, sets of two, three and four dependent variable were formed to reach most accurate regression number. So, two of models estimating dependent variables the most accurately were chosen in regression coeffients nearest to one. Selected models were applied to artificial nueral network progam to optimize tablet with requested criterias Requested criterias in selected models used to optimize tablets were breaking force of 9kg-f, releasing 50% of active substance in two hours and releasing 70% of active substance in five hours. Optimized tablets by modellin of these two formulas were tested. In control experiments almost 100% success in crushing strength, 91 % success in period to releas 50% of active substence and 96,4% success in amount of released active substence in five hours were achieved. Key words: Modelling, Optimized, Artificial Neural Network(ANN), diphenhidramine HCl and Tablet
Modelling and Optimization of Matrix tablets containing diphenhidramine HCl by using Artificial Neural Network In this study, diphenhidramine HCl as the active substance, an etanolamine derivative, and hydroxypropylmethyl cellulose (HPMC) polymer with two different viscosity, microcrystalline cellulose (Avicel) , lactose as excipients were used. The matrix tablets were prepared by direct compression method with hydrolic hand pressing equipment. Thirty two different tablet formulations were designed based on one master formula. The amount of active substence, diphenhidramine HCl of 50 mg and compression force of tablet 850kg-f were kept constant while the excipients were used in different amounts in sum of 325 mg. Different mixtures of excipients have changed tablet crushing strength, period to release 50% of the active substance and released active substence amount in fifth hour. In experiments, excipients HPMC100K, HPMC4K, Avicel and lactose were chosen independent variable while tablet crushing strength, period to release 50% of the active substance and released active substence amount in fifth hour were chosen as dependent variables. Then changes in dependent variables were observed. In order to maintain minumim side effects on dependent variables, mixing period and pressing force were kept constant on formulations. Controlled dependent variables in thirty two formulas were applied to the artificial neural network program for modelling. In modelling, sets of two, three and four dependent variable were formed to reach most accurate regression number. So, two of models estimating dependent variables the most accurately were chosen in regression coeffients nearest to one. Selected models were applied to artificial nueral network progam to optimize tablet with requested criterias Requested criterias in selected models used to optimize tablets were breaking force of 9kg-f, releasing 50% of active substance in two hours and releasing 70% of active substance in five hours. Optimized tablets by modellin of these two formulas were tested. In control experiments almost 100% success in crushing strength, 91 % success in period to releas 50% of active substence and 96,4% success in amount of released active substence in five hours were achieved. Key words: Modelling, Optimized, Artificial Neural Network(ANN), diphenhidramine HCl and Tablet