Issue link: https://www.e-digitaleditions.com/i/1543258
12 February 2026 Inhalation the time of writing this article, no AI- generated drug has yet received FDA approval, Insilico Medicine (Boston, Massachusetts, United States) recently pub- lished promising Phase IIa clinical trial results for its AI-identified candidate, Rentosertib, a novel serine/ threonine kinase (TNIK) inhibitor [24]. e results indicate that the drug is effective, well tolerated and suitable for Phase III trials. is may mark a pivotal moment, with an AI-identified drug candidate now advancing into later-stage clinical trials and approach- ing regulatory submission. Insilico Medicine's success is particularly relevant for the respiratory and inhaled drug development community, as Rentosertib is being trialed for the treatment of idiopathic pulmonary fibrosis, a degenerative lung condition with limited treatment options. Furthermore, Insilico Medicine has also announced the development of an inhalable formulation of its lead candidate [25]. AI is also used in other stages of drug development including formulation optimization, material char- acterization and clinical trial design [26, 27]. For example, AI-driven platforms are now being used to optimize formulation design, predict ideal compo- sitional ratios and refine release characteristics, such as identifying component combinations to achieve co-amorphous systems [28], moving beyond tradi- tional screening methods [21]. ese innovations may offer the potential to streamline development, reduce costs and minimize reliance on individual expertise, and could signal a future role in computa- tionally guided inhaled therapeutics [29, 30]. In silico modeling can also be used in inhaled drug development to predict particle behavior following inhalation. Computational simulations can play a role in understanding inhaled aerosol deposition in the human respiratory tract, particularly where direct experimental determination is challenging or unfea- sible. ese models can help predict the regional and total deposited fractions, and therefore the deposited dose, of inhaled particles within the lungs, which is vital for health risk assessment and aerosol therapy. Various modeling approaches exist, categorized by their structure and focus: whole-lung models, which can compute deposition across the entire lung, and local-scale models, which can provide detailed infor- mation on deposition patterns within specific lung regions like bronchial bifurcations [31]. Whole-lung models often use analytical equations for particle deposition efficiencies and specific flow conditions, while local-scale models solve particle transport and deposition equations using computational fluid and particle dynamics (CFPD) methods, offering insights into individual particle behavior [32, 33]. In silico approaches are also being used to simu- late the impact of formulation variables on aerosol performance. For example, models can assess how changes in excipient composition or propellant type influence aerosolization efficiency, droplet size and that changes will result in in vitro improvements or translate into clinical efficacy and efficiency. e high attrition rates of drug candidates observed in the late stages may be partly attributable to the limitations of utilizing purely empirical approaches in predicting product performance in vivo. Integrating computational tools with established experimental techniques may instead provide a promising path forward. Combining computational simulations with hands-on formulation expertise may offer a more data-directed and efficient devel- opment process, reducing reliance on experimentally guided lab work and may improve the probability of clinical success. In silico modeling: Applications in inhaled drug development In silico modeling is emerging as an important tool to support the increasingly complex task of develop- ing inhaled drug formulations. Leveraging compu- tational simulations can enable researchers to better predict, and therefore optimize, critical aspects of drug delivery, from the physicochemical properties of aerosols upon generation through their regional and total deposition in the lungs [18, 19]. ese tools can complement traditional experimental tech- niques by reducing reliance on extensive in vitro and in vivo testing, and may thereby accelerate develop- ment timelines and reduce costs. Computational methods are impacting early-stage drug development, particularly in the identification of novel drug candidates [20]. e ability to screen billions of molecules from extensive virtual libraries has increased the likelihood of discovering potent, selective compounds with favorable properties. In particular, structure-based virtual screening can enable accurate docking of molecules into 3D-recep- tor models, accelerating the identification of drug- like compounds and improving pharmacokinetic (PK) profiles early in development [21]. While these approaches are more established in systemic drug dis- covery, their direct application to inhaled medicines remains limited but holds considerable promise. However, challenges such as high computational costs, the need for extreme predictive accuracy and limited generalizability of some deep learning models, espe- cially when training data lacks diversity, continue to constrain broader adoption of computational meth- ods [21]. Nonetheless, the integration of computa- tional modeling is steadily progressing, with emerging technologies opening new avenues for inhaled drug development. In particular, the application of AI to pharmaceutical development is becoming increas- ingly prevalent and has significant scope to affect the industry [22]. A recent survey found that 80% of professionals in the pharmaceutical and life sciences sectors now use AI for drug discovery [23]. While at