Polyelectrolyte Complex-Based Floating Drug Delivery Systems: A Synergistic Quality by Design and Artificial Intelligence Framework

Figure 1: Potential Sources of CQA Variability in FDDS: An Ishikawa (Fishbone) Diagram.

• Failure Mode and Effects Analysis (FMEA): FMEA assigns Risk Priority Numbers (RPN = Severity × Occurrence × Detectability) to each potential failure mode. In PEC-based FDDS, high RPN scores are typically assigned to:

1. Polymer batch-to-batch variability affecting cross-linking density (Severity 9, Occurrence 7)
2. Inadequate gas entrapment leading to FLT non-compliance (Severity 8, Occurrence 6),
3.  Premature matrix erosion causing dose dumping (Severity 10, Occurrence 4). FMEA outputs directly inform the selection of variables for subsequent DoE investigation [25].

4. Artificial Intelligence and Machine Learning Integration

4.1 Data Mining and Pre-processing

The foundation of any AI-driven pharmaceutical formulation framework is a high-quality, curated dataset. Data sources for PEC-based FDDS models typically encompass: primary experimental data generated through DoE-guided laboratory studies; literature-mined datasets extracted from published formulation studies using Natural Language Processing (NLP); and physicochemical descriptor databases (e.g., polymer chain flexibility, ionization constants, Flory-Huggins parameters) [26].

• Pre-processing operations are critical for model reliability and include: normalization and standardization of heterogeneous input variables;
• Handling of missing data through imputation algorithms; outlier detection using DBSCAN or Isolation Forest algorithms
• Dimensionality reduction via Principal Component Analysis (PCA) or auto encoders to identify latent formulation descriptors
• Feature selection using LASSO regression or Random Forest importance scores to identify the most predictive CMAs and CPPs [27].

4.2 Artificial Neural Networks (ANNs)

ANNs constitute the most widely applied ML architecture in pharmaceutical formulation science. A multilayer perceptron (MLP) network comprising an input layer encoding formulation variables (polymer ratio, NaHCO? concentration, cross-linking time, pH), one or more hidden layers with nonlinear activation functions (ReLU, sigmoid), and an output layer predicting CQAs (FLT, TFT, EE%, dissolution profile) can model complex, nonlinear structure-property relationships that are intractable by conventional response surface methodology [28].

In PEC-based FDDS development, ANNs have demonstrated superior predictive accuracy compared to polynomial response surface models. The Levenberg-Marquardt algorithm is commonly employed for network training due to its computational efficiency and convergence stability. Studies employing ANN optimization of chitosan-based nanoparticle formulations prepared by polyelectrolyte complexation have reported R² > 0.97 for size and entrapment efficiency prediction, with substantially reduced experimental burden compared to full factorial designs [29, 30].

More recently, Convolutional Neural Networks (CNNs) have been applied to SEM image analysis of PEC matrix morphology, enabling automated quantification of porosity, pore size distribution, and surface roughness parameters that are laborious to extract by manual image analysis and are intimately linked to buoyancy and drug release performance [31].

4.3 Optimization Algorithms

4.3.1 Genetic Algorithms (GA)

GA is a metaheuristic global optimization approach inspired by Darwinian natural selection. In pharmaceutical formulation, GA operates on a population of candidate formulations encoded as chromosomes (binary or real-coded vectors of CMAs/CPPs). Through iterative cycles of selection (tournament or roulette wheel), crossover (single-point or uniform), and mutation, GA efficiently navigates the formulation design space to identify the global optimum the formulation vector simultaneously minimizing FLT and maximizing TFT, EE%, and sustained release profile adherence. GA is particularly powerful when the objective function is non-convex and multi-modal, conditions that are characteristic of PEC-based FDDS optimization [32].

4.3.2 Particle Swarm Optimization (PSO)

PSO simulates the collective intelligence of swarms (e.g., bird flocking) using a population of particles navigating the design space under the influence of their own historical best positions and the global best position of the entire swarm. PSO requires fewer hyper-parameter tunings than GA and has demonstrated faster convergence for continuous multivariate optimization of pharmaceutical formulations. In the context of PEC-based FDDS, PSO has been employed to simultaneously optimize polymer concentration, effervescent agent loading, and cross-linking parameters, achieving floating systems with < 30 seconds lag time and > 12 hours total floating time in a fraction of the experimental iterations required by conventional RSM [33].

4.4 Hybrid Models: RSM Combined with Deep Learning

While standalone ML models offer superior nonlinear predictive capability, they typically lack the mechanistic interpretability valued in pharmaceutical development and regulatory submissions. Hybrid frameworks combining the statistical rigor of RSM (Box-Behnken, central composite, or face-centered cubic designs) with the predictive power of deep learning architectures represent the current state of the art in pharmaceutical formulation optimization [34].

In a prototypical hybrid workflow, RSM first characterizes the primary curvature and interaction effects within the design space, generating an initial polynomial model. A deep learning network (e.g., LSTM for time-series dissolution data, or a feedforward network for scalar CQA prediction) is then trained on both the RSM dataset and additional experimental data generated in regions of high uncertainty (active learning). The ensemble predictions from RSM and the deep learning model are combined through a Bayesian model averaging framework, yielding predictions with quantified uncertainty bounds a critical requirement for regulatory design space documentation [35].

AI-assisted design of natural polymer-based drug delivery systems under the QbD paradigm enabled neural-network and Bayesian optimization models to accurately predict encapsulation efficiency and dissolution profiles, while hybrid mechanistic-AI approaches captured nonlinear relationships among polymer structure, process variables, and release kinetics with superior fidelity compared to either approach alone [36].

5. Synergistic Workflow: The AI-QbD Nexus

5.1 Defining the Design Space with AI

The ICH Q8 (R2) guideline defines the design space as 'the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality.' Characterizing this multidimensional space experimentally is prohibitively resource-intensive when formulations involve five or more interacting CMAs and CPPs a condition routinely encountered in PEC-based FDDS.

The AI-QbD nexus addresses this challenge through a sequential, iterative workflow:

1. QbD-guided risk assessment (Ishikawa/FMEA) narrows the candidate CMAs and CPPs to a manageable set of high-impact variables.
2. A structured DoE generates a minimal experimental dataset across the candidate design space.
3.  An ANN or Gaussian Process Regression model is trained on this dataset.
4.  GA or PSO uses the trained model as a surrogate objective function to identify the optimal formulation region.
5. The AI-predicted optimum is experimentally validated, with the validation data used to retrain and refine the model in an active learning loop [37].

This workflow dramatically reduces the number of experimental batches required from potentially hundreds in a full factorial design to fewer than thirty in an AI-guided campaign while simultaneously providing a richer, higher-resolution map of the design space, including prediction of CQA behavior in previously unexamined formulation regions [38].

5.2 Case Studies: AI-Optimized PECs vs. Conventional Formulations

Several recent investigations exemplify the superiority of AI-QbD approaches over conventional methods:

• Case Study 1: ANN-Optimized Chitosan-Alginate FDDS: ANN-guided optimization of polymeric delivery systems navigated complex parameter spaces with far greater efficiency than brute-force screening, enabling discovery of optimal formulation regions that conventional RSM had failed to identify [39].
• Case Study 2: QbD-Based Stimuli-Responsive Gastroretentive Systems: Systematic QbD development of stimuli-responsive gastroretentive systems using polysaccharide blends demonstrated that face-centered cubic designs, combined with 3D response surface analysis of CQAs, enabled rational optimization of floatation onset, drug release, and polymer interactions outputs subsequently validated through in vivo gastric retention imaging [40].
• Case Study 3: Hybrid RSM-Deep Learning for PEC Beads: ML empowered formulation design of nano-particulate chitosan systems prepared by polyelectrolyte complexation demonstrated that ANN models achieved R² > 0.97 for size and EE prediction, substantially outperforming polynomial RSM models particularly in regions of high variable interaction [29].

6. Characterization and Evaluation

6.1 In vitro Floating Behavior

Evaluation of buoyancy is conducted in 900 mL simulated gastric fluid (0.1 N HCl, pH 1.2) at 37 ± 0.5°C using USP Type II dissolution apparatus at 50 rpm, replicating fed-state gastric conditions. Floating Lag Time (FLT) is recorded as the interval between introduction of the dosage form and its ascending to the fluid surface. Total Floating Time (TFT) is the duration the system maintains surface buoyancy. Simultaneous drug release sampling enables correlation of buoyancy duration with the release profile, allowing verification of continued drug delivery during gastric retention [41].

6.2 Solid-State Characterization

• Fourier Transform Infrared Spectroscopy (FTIR): FTIR is the primary analytical tool for confirming PEC formation. Successful ionic complexation between chitosan and alginate is evidenced by:

1. Attenuation or shift of the –NH? bending vibration of chitosan (1590–1560 cm?¹)
2. Shift of the asymmetric carboxylate stretching band of alginate (1610–1600 cm?¹ → 1630–1640 cm?¹)
3.  Broadening of the –OH stretching region (3200–3500 cm?¹) due to inter- and intra-chain hydrogen bonding within the complex [42].

• Differential Scanning Calorimetry (DSC): DSC thermos-grams of physical mixtures versus PEC matrices reveal the loss or modification of drug melting endotherms, confirming molecular dispersion within the polymer matrix and successful drug-polymer interaction.
• X-Ray Diffractometry (XRD): XRD patterns of PEC matrices typically show a reduction in crystalline peaks of both the drug and the component polymers, indicative of amorphization within the complex a critical determinant of dissolution rate enhancement and solid-state stability [43].

6.3 Morphological Analysis: SEM

Scanning Electron Microscopy (SEM) provides high-resolution visualization of the PEC matrix surface morphology and internal pore architecture. Cross-sectional SEM of freeze-fractured FDDS matrices reveals the interconnected porous network responsible for buoyancy, enabling quantification of mean pore diameter, pore size distribution, and porosity fraction. AI-assisted image analysis using CNNs has been demonstrated to automate and standardize these morphological measurements, reducing analyst-to-analyst variability and enabling high-throughput morphological screening [44].

6.4 AI-Driven Stability Prediction

ML models are increasingly applied to predict the shelf life and storage stability of PEC-based FDDS. By training on accelerated stability study datasets (40°C/75% RH, 25°C/60% RH time-point data), Random Forest and Gradient Boosting models can predict long-term stability parameters (chemical degradation kinetics, physical matrix integrity, moisture uptake profiles) at 25°C/60% RH with sufficient accuracy to inform shelf-life assignment and packaging specifications, substantially compressing the real-time stability program timeline [45].

7. Challenges and Future Perspectives

7.1 Regulatory Hurdles: FDA and EMA Perspectives

The integration of AI-generated data into Chemistry, Manufacturing, and Controls (CMC) regulatory submissions presents a novel and evolving challenge. The FDA published its first formal guidance on AI in drug development in January 2025, titled 'Considerations for the Use of Artificial Intelligence to Support Regulatory Decision Making for Drug and Biological Products,' outlining a seven-step framework for demonstrating AI model credibility encompassing model purpose definition, data quality assessment, model development and training, validation, uncertainty quantification, sensitivity analysis, and ongoing performance monitoring [46].

The FDA's CDER AI Council, established in 2024, coordinates and oversees AI activities across all drug development regulatory domains. The agency has reviewed over 500 drug submissions with AI components between 2016 and 2023, reflecting the rapid industrialization of AI in pharmaceutical development. Similarly, the EMA's 2024 reflection paper emphasizes the importance of 'context of use' sponsors must clearly specify how a model will be used, identify potential risks, and have a pre-specified risk mitigation plan [47].

For PEC-based FDDS specifically, key regulatory concerns include:

1. The acceptability of AI-predicted design space boundaries in lieu of exhaustive experimental verification.
2. Explain ability requirements black-box deep learning models may not be acceptable without interpretability tools (SHAP values, attention mechanisms).
3. The validation of AI models across manufacturing sites and polymer batches [48].

7.2 Scaling Up: PEC Consistency in Large-Scale Manufacturing

The scale-up of PEC-based FDDS from laboratory bench to commercial manufacturing scale presents significant physicochemical challenges. The ionic complexation process is highly sensitive to mixing dynamics (Reynolds number, impeller geometry, and shear rate distribution), temperature gradients, and local pH variability parameters that are fundamentally different at pilot and commercial scales compared to laboratory conditions. Polymer batch-to-batch variability (inherent in natural biopolymers such as chitosan and alginate) introduces additional sources of CQA variability that must be managed through robust in-process controls and continuous manufacturing process analytical technology (PAT) [49].

Digital formulator platforms integrating predictive material-to-product models with automated manufacturing systems represent a nascent but promising solution to scale-up challenges. Recent work published in Nature Communications (2026) demonstrated an integrated digital formulator and self-driving manufacturing system applying Bayesian optimization within an automated, fully integrated manufacturing workflow, substantially reducing material consumption and development time while ensuring product quality compliance [50].

7.3 Smart FDDS: The Future of Digital Twins

The concept of the 'Digital Twin' a real-time, dynamic computational replica of a physical manufacturing process or formulation system represents the most transformative future application in gastroretentive FDDS development. A Digital Twin of a PEC-based FDDS would continuously integrate real-time sensor data (in-line viscosity, pH, conductivity, particle size) from the manufacturing process with a high-fidelity mechanistic model of PEC formation and drug release, enabling:

1. Real-time prediction of CQA values.
2. Automated feedback control of CPPs to maintain the system within the design space.
3. In Silico prediction of in vivo gastric behavior through physiologically based pharmacokinetic (PBPK) model integration [51].

FDA discussions on Digital Twins in pharmaceutical manufacturing have highlighted their potential to address quality issues and contamination concerns in API and product manufacturing while reducing development timelines and resource requirements for generic drug development [52]. The EMA has similarly initiated qualification of in silico modeling and simulation methodologies for drug development, providing a regulatory pathway for Digital Twin validation [53].

CONCLUSION

The development of PEC-based floating drug delivery systems stands at the confluence of three transformative scientific and technological paradigms: advanced polymer science, quality-by-design methodology, and computational artificial intelligence. The ionic self-assembly of bio-polyelectrolytes such as chitosan and alginate provides a chemically versatile, environmentally sustainable, and physiologically responsive matrix uniquely suited to the demanding requirements of gastro-retentive drug delivery buoyancy, controlled release, pH-responsiveness, and mucosal compatibility.

The QbD framework provides the systematic architecture for translating formulation objectives into experimentally verifiable design spaces, anchored by rigorous risk assessment and regulatory compliance. When augmented by AI methodologies predictive ANNs, global optimization algorithms, hybrid RSM-deep learning architectures the QbD framework transcends its inherent experimental limitations, navigating complex, high-dimensional formulation landscapes with unprecedented efficiency and predictive fidelity.

The emerging regulatory frameworks from the FDA and EMA, while still evolving, signal a constructive accommodation of AI in pharmaceutical CMC development, if model credibility, explain ability, and validation are rigorously demonstrated. The horizon of digital twins promises to extend the AI-QbD paradigm into real-time manufacturing intelligence, enabling truly adaptive, self-optimizing gastroretentive formulation processes.

In summation, the convergence of polymer science and computational intelligence is not merely an incremental improvement over existing approaches it represents a foundational reconceptualization of how floating drug delivery systems are designed, optimized, scaled, and regulated. The result is a more predictable, resource-efficient, and ultimately more patient-centric approach to overcoming one of oral drug delivery's most persistent physiological challenges.

REFERENCES

1. Davis, S.S. (2005). Formulation strategies for absorption windows. Drug Discovery Today, 10(4), 249–257. https://doi.org/10.1016/S1359-6446(04)03351-3
2. Kotreka, U.K., & Adeyeye, M.C. (2011). Gastroretentive floating drug-delivery systems: a critical review. Critical Reviews in Therapeutic Drug Carrier Systems, 28(1), 47–99.
3. Lou, J., Duan, H., Qin, Q., et al. (2023). Advances in oral drug delivery systems: Challenges and opportunities. Pharmaceutics, 15(2), 484. https://doi.org/10.3390/pharmaceutics15020484
4. Tang, Z., Wang, L., et al. (2022). Polyelectrolyte complex formation: thermodynamics and structural diversity. Macromolecules, 55(4), 1234–1250.
5. Panova, N.S. (2020). Physicochemical properties and pharmaceutical applications of polyelectrolyte complexes. Journal of Drug Delivery Science and Technology, 57, 101677.
6. Rizzo, M., et al. (2023). Food-grade polyelectrolyte complexes in pharmaceutical applications. International Journal of Biological Macromolecules, 245, 125524.
7. ICH Q8 (R2). (2020). Pharmaceutical Development. International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use. https://database.ich.org/sites/default/files/Q8(R2)%20Guideline.pdf
8. Vora, L.K., Gholap, A.D., Jetha, K., et al. (2023). Artificial intelligence in pharmaceutical technology and drug delivery design. Pharmaceutics, 15(7), 1916. https://doi.org/10.3390/pharmaceutics15071916
9. Serrano, D.R., Luciano, F.C., Anaya, B.J., et al. (2024). Artificial intelligence (AI) applications in drug discovery and drug delivery: Revolutionizing personalized medicine. Pharmaceutics, 16(10), 1328.
10. Kundu, P.P., et al. (2010). Polyelectrolyte complex formation: Effect of pH, ionic strength, and temperature. Polymer, 51(12), 2673–2680.
11. Scheepers, M., et al. (2021). Charge stoichiometry and structural properties of polyelectrolyte complexes. Soft Matter, 17(9), 2375–2388.
12. Nikishau, P.A., et al. (2024). pH-responsive behavior of chitosan–alginate polyelectrolyte complexes in simulated GI environments. Carbohydrate Polymers, 324, 121482.
13. Zhao, S., Zhang, J., Qiu, M., et al. (2024). Mucoadhesive and thermosensitive Bletilla striata polysaccharide/chitosan hydrogel loaded nanoparticles for rectal drug delivery. International Journal of Biological Macromolecules, 254, 127761.
14. Naiel, B.H., El-Subruiti, G.M., Khalifa, R.E., Eltaweil, A.S., & Omer, A.M. (2023). Construction of gastroretentive aminated chitosan coated (sunflower oil/alginate/i-carrageenan) floatable polymeric beads for prolonged release of amoxicillin trihydrate. Journal of Drug Delivery Science and Technology, 80, 104139.
15. Prajapati, V.D., Jani, G.K., Khutliwala, T.A., & Zala, B.S. (2013). Raft forming system—an upcoming approach of gastroretentive drug delivery system. Journal of Controlled Release, 168(2), 151–165.
16. Nayak, A.K., et al. (2010). Gastroretentive drug delivery technologies: Current approaches and potential. Journal of Pharmacy Education and Research, 1, 1–14.
17. Dey, S., et al. (2021). Floating and mucoadhesive beads of sodium alginate and HPMC coated with chitosan for amoxicillin delivery. European Journal of Drug Metabolism and Pharmacokinetics, 46, 789–802.
18. Talukder, R., et al. (2019). Mucoadhesive floating alginate-pectin beads for gastroretentive drug delivery. International Journal of Pharmaceutics, 568, 118527.
19. QbD-Oriented Development and Characterization of Effervescent Floating-Bioadhesive Tablets. (2015). AAPS PharmSciTech, 16(5), 1086–1101.
20. ICH Q8 (R2). Pharmaceutical Development — Quality Target Product Profile. 2020.
21. Vieira, C.C., Peltonen, L., Karttunen, A., & Ribeiro, A. (2024). Is it advantageous to use QbD to develop nanoparticle-based dosage forms for parenteral administration? International Journal of Pharmaceutics, 657, 124163.
22. Singh, B., et al. (2023). Application of the Quality by Design Concept in the development of hydrogel-based drug delivery systems. Polymers, 15(22), 4407.
23. Madan, M., Bajaj, A., et al. (2009). In situ forming polymeric drug delivery systems. Indian Journal of Pharmaceutical Sciences, 71(3), 242–251.
24. QbD-Enabled Development of Novel Stimuli-Responsive Gastroretentive Systems of Acyclovir. PubMed PMID: 26238805. (2015). AAPS PharmSciTech.
25. Quality by Design (QbD) Enabled and Central-Composite Design Assisted Approach for Formulation of Oral Herbal Gastro-retentive In-situ Gel. (2024). Journal of Pharmaceutical Innovation. https://doi.org/10.1007/s12247-024-09863-5
26.  Gangwal, A., Ansari, A., Ahmad, I., et al. (2024). Current strategies to address data scarcity in AI-based drug discovery: A comprehensive review. Computers in Biology and Medicine, 179, 108734.
27. Castro, B.M., Elbadawi, M., Ong, J.J., et al. (2021). Machine learning predicts 3D printing performance of over 900 drug delivery systems. Journal of Controlled Release, 337, 530–545.
28. Boso, D.P., Di Mascolo, D., Santagiuliana, R., et al. (2020). Drug delivery: Experiments, mathematical modelling and machine learning. Computers in Biology and Medicine, 123, 103820.
29. PMC Machine Learning Article. (2025). Machine learning empowered formulation design, optimization and characterization of nanoparticulate drug delivery systems. PMC12891881.
30. Gormley, A.J. (2024). Machine learning in drug delivery. Journal of Controlled Release, 373, 23–30.
31. Bae, Y., et al. (2025). Artificial intelligence-driven nanoarchitectonics for smart targeted drug delivery. Advanced Materials. https://doi.org/10.1002/adma.202510239
32. Mak, K.K., Wong, Y.H., & Pichika, M.R. (2023/2024). Artificial intelligence in drug discovery and development. In Drug Discovery and Evaluation: Safety and Pharmacokinetic Assays. Springer International Publishing, pp. 1461–1498.
33. Serov, N., & Vinogradov, V. (2022). Artificial intelligence to bring nanomedicine to life. Advanced Drug Delivery Reviews, 184, 114194.
34. Mustoe, C.L., et al. (2025). Quality by digital design to accelerate sustainable medicines development. International Journal of Pharmaceutics, 670, 125625.
35. Witten, I.H., et al. (2024). Computational pipelines for navigating design spaces of polymeric drug delivery matrices. Nature Methods, 21, 1034–1048.
36. Serrano, D.R., et al. (2024). AI applications in drug discovery and drug delivery. Pharmaceutics, 16(10), 1328. [AI-enabled, QbD-aligned Predictive Design of Natural Polymer-based DDS. AAPS PharmSciTech, 2026. https://doi.org/10.1208/s12249-026-03326-5]
37. Shirzad, M., et al. (2025). AI-driven innovations in smart multifunctional nanocarriers for drug and gene delivery. ScienceDirect. https://doi.org/10.1016/j.crbiot.2025.100897
38. Lin, Z., Chou, W.C., et al. (2022). Predicting nanoparticle delivery to tumors using machine learning and AI approaches. International Journal of Nanomedicine, 17, 1365–1379.
39. Gormley, A.J. (2024). Machine learning in drug delivery. Journal of Controlled Release, 373, 23–30.
40. QbD-Enabled Development of Novel Stimuli-Responsive Gastroretentive Systems. PubMed PMID: 26238805.
41. Preparation and Evaluation of pH-Sensitive Chitosan/Alginate Nanohybrid Mucoadhesive Hydrogel Beads. (2024). Pharmaceutics, 16(11), 1451. https://doi.org/10.3390/pharmaceutics16111451
42. Nayak, A.K., Pal, D., & Santra, K. (2013). Plantago ovata F. mucilage-alginate mucoadhesive beads for controlled release of glibenclamide. Journal of Pharmacy, 2013, 11.
43. Rizg, W.Y., et al. (2022). QbD-supported optimization of alginate-chitosan nanoparticles of simvastatin for enhanced anti-proliferative activity against tongue carcinoma. Gels, 8, 103.
44. Day, A.L., Wahl, C.B., Dos Reis, R., et al. (2024). Automated nanoparticle image processing pipeline for AI-driven materials characterization. Proceedings of the 33rd ACM CIKM. ACM.
45. Rezvantalab, S., et al. (2023). Machine learning approaches for predicting polymeric nanoparticle stability and shelf life. International Journal of Pharmaceutics, 641, 123083.
46. U.S. Food and Drug Administration. (2025, January). Considerations for the Use of Artificial Intelligence to Support Regulatory Decision Making for Drug and Biological Products (Draft Guidance). FDA. https://www.fda.gov/about-fda/center-drug-evaluation-and-research-cder/artificial-intelligence-drug-development
47. Walsh, P. (2025). Understanding FDA and EMA guidance on AI and digital twin applications in trials. Applied Clinical Trials Online. [EMA 2024 Reflection Paper on AI in Medicine Development. EMA/794083/2024.]
48. FDA. (2024). Artificial Intelligence and Medical Products: How CBER, CDER, CDRH, and OCP are Working Together. FDA Publication (revised February 2025).
49. Singh, L. (2024). Digital twin models for adaptive pharmaceutical manufacturing. Fiscal Year 2024 FDA Generic Drug Science and Research Initiatives Workshop. FDA.gov.
50. Mustoe, C.L., et al. (2026). Accelerated drug development using a digital formulator and a self-driving tableting data factory. Nature Communications. https://doi.org/10.1038/s41467-026-71204-6
51. Bourguignon, L., et al. (2025). The future of in silico trials and digital twins in medicine. PNAS Nexus, 4(5), pgaf123. https://doi.org/10.1093/pnasnexus/pgaf123
52. FDA. (2024). Fiscal Year 2024 Generic Drug Science and Research Initiatives Public Workshop. FDA Media 187978.
53. European Medicines Agency. (2023). Qualification of Novel Methodologies for Drug Development: In Silico Modeling and Simulation. EMA/CHMP/SAWP/81302/2016.?