In silico models for skin permeability prediction – Alternative ways to replace animal testing

  • Xin Ling Quah

Student thesis: Doctoral Thesis

Abstract

Skin permeation plays an important role in drug administration for transdermal and topical products in the pharmaceutical industry. To develop in silico models for skin permeability prediction, a freely accessible database ‘HuskinDB’ which focuses purely on human skin that has recently been released was chosen. After consideration of the database selection, molecular descriptors were analysed in the study to decide which descriptors should be included for model development. In this research, two different types of molecular descriptors: physicochemical properties and functional groups were examined separately for model development. Firstly, partition coefficient (Log P), topological polar surface area (TPSA) and molecular volume (MV) were chosen as the three main molecular descriptors for model development based on physicochemical properties. An optimum skin permeability prediction model of LogKp = 0.291 Log P − 0.001 TPSA − 0.005 MV − 5.775, with 214 compounds was then achieved by QSPR analysis with a correlation coefficient of 0.5044. Secondly, ten functional groups: ketone, carboxylic acid, hydroxyl, amine, ether, aromatic, amide, ester, chlorine and bromine groups were selected as the final molecular descriptors for model development utilising functional groups. A skin permeability prediction model of LogKp = − 0.221 Ketone − 0.595 Carboxylic acid − 0.249 Hydroxyl − 0.186 Amine − 0.202 Ether + 0.133 Aromatic − 0.350 Amide − 0.309 Ester + 0.169 Chlorine + 0.488 Bromine − 5.882, with 180 compounds was established with a correlation coefficient of 0.5030. This latter approach, i.e. utilising functional groups for prediction was particularly novel. Finally, an investigation was conducted on two additional ADME properties: blood-brain barrier (BBB) permeation and human gastrointestinal absorption (HIA) to determine whether this approach was suitable for other predictive systems. In summary, this study has confirmed that skin permeability can be predicted using both physicochemical properties and functional groups. Both variables can predict the overall skin permeability (LogKp), but prediction using functional groups is truly unique and simply from the chemical structure of the compound. Beyond skin permeability, the novel application of functional groups for in silico model development can also be utilised for other ADME predictions as proposed in this work with models created for blood-brain barrier (BBB) permeation and human gastrointestinal absorption (HIA).
Date of Award16 Jan 2024
Original languageEnglish
SupervisorLaura Waters (Main Supervisor) & Tianhua Chen (Co-Supervisor)

Cite this

'