ICDSUPL1-T007

Volume: 1, 2022
1^st International PhD Student’s Conference at the University of Life Sciences in Lublin, Poland: ENVIRONMENT  – PLANT  – ANIMAL  – PRODUCT

Abstract number: T007

DOI: https://doi.org/10.24326/ICDSUPL1.T007

Published online: 26 April 2022

ICDSUPL, 1, T007 (2022)

Analysis of liquisolid systems’ tableting performance by machine-learning algorithms

Teodora Glisic¹*, Jelena Djuris¹, Ivana Vasiljevic¹, Ivana Aleksic¹

^{1 Department of Pharmaceutical Technology and Cosmetology, University of Belgrade – Faculty of Pharmacy, Vojvode Stepe 450, 11221 Belgrade, Serbia.}

^{* Correspondence author: teodora.glisic@hotmail.com}

Abstract

Liquisolid systems (LSS) have been recognized as a promising approach in improving oral bioavailability of poorly soluble drug substances. However, tableting of these formulations has proven to be challenging since it requires both good flowability and compaction properties while retaining relatively high amount of the liquid phase. These properties are influenced by a number of variables, such as characteristics of the excipients used, formulation factors as well as some parameters specific to the tableting process. The aim of this study was to apply machine-learning classification and regression algorithms for better understanding of the relationship between the type of carrier used, liquid load, carrier/coating material ratio and tableting process variables and compact tensile strength. Compression pressure, detachment and ejection stress, net work of compression and elastic recovery were tableting process parameters used for the model development. The analyzed data set consisted of 8 numeric and 1 categorical variable with 150 entries each. During the preliminary analysis the effect of the type of carrier used, liquid load and carrier/coating material ratio on flowability was analyzed as well. The results of regression analysis showed that AdaBoost algorithm was the best fitted model for tablet tensile strength as a target variable (coefficient of determination=0.95), with the ejection stress, the parameter that can be monitored during the tableting process, being the input variable with the greatest influence. The model was also affected by the type of carrier used as well as compression pressure. Classification analysis indicated that AdaBoost algorithm could also be used as a predictive model (precision=0.901) to cluster the data according to the type of carrier used with detachment stress, ejection stress and tensile strength being the most significant variables. It was noted that LS admixtures’ flowability was mostly affected by the type of carrier used, which could be related to differences in particle size and shape, and by liquid load. Machine learning algorithms allow a more detailed and comprehensive overview of the complex relationships between different factors that affect characteristics of LS formulations and could be a useful tool in their development and manufacturing.

How to cite

T. Glisic, J. Djuris, I. Vasiljevic, I. Aleksic, 2022. Analysis of liquisolid systems’ tableting performance by machine-learning algorithms. In: 1st International PhD Student’s Conference at the University of Life Sciences in Lublin, Poland: Environment – Plant – Animal – Product. https://doi.org/10.24326/ICDSUPL1/T007