Drug delivery via the gastrointestinal tract remains the dominant route for administration of many medicinals. However, when delivered this way, such bioactive chemicals are susceptible to gastrointestinal enzymatic degradation and extensive first-pass metabolism in the liver that limit their bioavailability. Parenteral administration circumvents these issues but introduces problems with compliance, particularly with the young, those with trypanophobia (fear of needles), and often requires clinical supervision to dispense medication. Transdermal delivery is a possibility but is challenging for most medicinals because of the low permeability of the skin. Not only is the skin’s epidermis highly keratinised (stratum corneum), but it contains a significant permeability barrier in the upper granular layers of the stratum spinosum, just below the stratum corneum. Here, secreted ceramides, fatty acids and cholesterol are organised into lipid-rich lamellae and gel-phase that fill the intercellular spaces around corneocytes, producing a highly hydrophobic layer that is largely impermeable to all but lipophilic chemicals (Proksch et al., 2008). Epidermal permeability is further regulated by tight junctions present on the plasma membrane of keratinocytes in the stratum spinosum. Tight junctions are comprised of a complex of proteins that include claudins, occludin, junctional adhesion molecules (JAM) and zona occludins (ZO), which act by pulling adjacent cells tightly together, forming a gate that regulates the passage of molecules to the deeper layers of the epidermis and into the dermis (Otani and Furuse, 2020).

The difficulties associated with skin-mediated drug delivery has led to advances in smart-material devices such as mucoadhesive tablets, cast films or electrospun membranes for topical administration directly to or through the oral mucosa (tissues lining the mouth), particularly the buccal (cheek) or sublingual (under the tongue) mucosa (Liu et al., 2025, Malhotra et al., 2025, Edmans et al., 2020, Bahraminejad and Almoazen, 2025, Mauceri et al., 2025). The attraction for delivering medicinals via the oral mucosa rather than skin is the difference in permeability between these two types of epithelia, with the oral mucosa possessing permeability that is several times greater than the skin (Galey et al., 1976). This is because the buccal/sublingual epithelium is non-keratinised (devoid of a stratum corneum) and contains very few lamellae and membrane-coating granules (Hashimoto et al., 1966), meaning that its lipid-rich permeability barrier is markedly reduced compared to skin (Proksch et al., 2008), allowing many chemicals to readily permeate this type of epithelium. The delivery of several chemicals and even small peptides has been reported by these mucosal devices but as of yet, only a few (e.g., ondansetron, buprenorphine, naloxone, fentanyl, nicotine) are available on the market (Malhotra et al., 2025, Zhang et al., 2002).

Depending on their physiochemical properties (such as lipophilicity, charge, molecular weight and geometry etc) chemicals traverse the oral non-keratinised epithelium either across (transcellular) or between (paracellular) oral keratinocytes. Generally, small lipophilic chemicals favour the transcellular permeation route as they can dissolve in the amphiphilic plasma membranes of keratinocytes and in the lipid-rich lamellae. The low levels of lipids within the outer stratum spinosum layers of buccal/sublingual epithelium mean that tight junctions are the principal determinants regulating the permeation of more hydrophilic molecules through the paracellular space in this tissue (Samiei et al., 2019).

Due to technological advancements in drug delivery, it is projected that the number of chemicals delivered via the buccal/sublingual mucosa will dramatically increase (Bahraminejad and Almoazen, 2025). In addition, there is also a need to assess the toxicity profiles of other agents, such as agrochemicals and oral health products that come into contact with the oral mucosa. Laboratory-based methods to measure chemical permeation include the use of porcine buccal mucosa or tissue engineered in vitro human buccal mucosa that are time consuming and would need to be performed for every chemical screened. An alternative approach is to pre-screen using in silico modelling to predict chemical permeability.

In silico models for tissue permeation are commonly either compartmental (Xia et al., 2015, Polak et al., 2012) or diffusion-based (Kruse et al., 2007, Verma et al., 2023) that are inexpensive to compute but experience low-fidelity. Compartmental models assume that the chemical concentration in a tissue is homogeneous and governed by an ordinary differential equation in time; physiologically-based pharmacokinetic (PBPK) models generally rely on these for whole-body predictions (Peters, 2008). Diffusion-based models incorporate spatial variation in chemical concentration but rely on a diffusion coefficient which often represents a homogeneous tissue. Spatially varying diffusion coefficients are possible but any local discontinuities, such as cell boundaries cannot be accounted for directly. Therefore, in both compartmental and diffusive models, each tissue or organ is described by macroscale parameters that must be fitted or approximated. Along with models fitted with machine learning (Biswas et al., 2024, Sun et al., 2011), neither are viable for accurate predictions of tissue permeation unless very large datasets are obtained.

In contrast, high-fidelity cell-based models can capture the inherent inhomogeneity of a tissue directly, with existing examples representing hepatocyte spheroids (Leedale et al., 2020) and in vitro oral mucosal tissue (Edwards et al., 2024). Chemical permeation is modelled mechanistically, with diffusion through paracellular routes and/or permeation of cell membranes and transcellular pathways. In particular, the chemical concentration in the tissue is modelled with partial differential equations in multiple spatial dimensions. Whilst cell-based models provide obvious advantages, their adoption has been limited due to their inherent complexity and the necessity of greater computational resources. However, these limitations are diminishing with the availability and affordability of increasing computational power.

Using tissue engineered human buccal mucosa, we previously developed an advanced mechanistic mathematical model of the buccal epithelium (Edwards et al., 2024). This was used to create a histologically and physiologically-relevant in silico model of buccal mucosal chemical permeation using partial differential equations, fitted to chemical permeation from in vitro assay data for chemicals with known physiochemical properties (Edwards et al., 2024). Our previous study also elucidated, for the first time, the influence of convoluted extracellular space on epithelial drug permeation, which was also factored into the in silico model, making this a significant improvement on previous models where chemical permeation is likely underpredicted (Edwards et al., 2024). Moreover, our in silico model could be used to make perturbations in, for example, the size of the extracellular spaces to mimic presence of epithelial permeation enhances where this predicted the enhanced permeation of chemicals that pass via the paracellular route (Edwards et al., 2024). However, this model was based on data generated from tissue engineered in vitro buccal mucosal studies. Structural differences between tissue engineered in vitro and in vivo human tissues, such as size and the number of layers of the keratinocytes in the stratum spinosum (Edwards et al., 2024), necessitates the development of a mathematical in silico model that can map from in vitro to in vivo, to enable human-specific chemical permeation predictions.

This current study aims to overcome these problems by generating novel tissue engineered buccal mucosa with reduced permeability barrier properties to better model the buccal mucosa chemical permeability barrier. We further developed a new in silico method that maps in vitro studies to in vivo prediction, where human in vivo buccal tissue and cell geometry were included directly so that in vitro and in vivo differences were accounted for. Furthermore, the dual impact of passive membrane transport and the buccal permeability barrier was ascertained. With in vivo predictions established, the blood vessel density of the buccal mucosa was assessed using live in vivo imaging, and the predictions from the in silico model fed into a PBPK model to map whole body chemical distribution. This novel integrated pipeline supports extrapolation from in vitro to in vivo, potentially improving future compound screening and formulation design for chemicals permeating the buccal mucosa, while reducing reliance on animal tissue.