Oral NDA Research from AI Prediction to Clinical Drug Formulation
Joe Chou1, Roger Lai2, Jason Chou1, Shelly Fu2, Benjamin Chang1
14th Floor, No. 366-1, Jilong Road Section 1, Taipei, Taiwan.
26F-11, No. 26, Tai-Yuen Street, ChuPei, Hsinchu, Taiwan.
*Corresponding Author E-mail: a0968288562@gmail.com
ABSTRACT:
Recent advancements in AI drug prediction have dramatically improved new drug candidate screening, facilitating more efficient identification of potential compounds. The success of a new drug application (NDA) relies on early-stage drug prediction and screening, preclinical studies，and PK/PD simulation prior to clinical study, underscoring the importance of accurate predictive models. Several essential databases, like Protein Data Bank (PDB), OpenTarget, PubChem, and ChemBL, SwissTarget etc., have made new drug searches easier with the help of fast AI computation, streamlining the drug discovery process. In drug dosage design, the oral formulation is still the most commonly needed due to its convenience in administration, emphasizing the significance of optimizing oral drug formulations for efficacy and patient compliance. Recent reports in oral new drug development have pointed out that PBPK modeling may lead to better prediction of ADME, offering a promising avenue for enhancing the understanding of drug behavior in the body. In addition, an alternative method using a FDA-approved PK database and PAMPA dissolution is also proposed to improve the development of NDA oral formulation, providing complementary approaches for formulation optimization. Therefore, a better prediction of an NCE from AI screening to clinical drug formulation could be conducted to enhance the success rate of oral NDA formulation development, fostering advancements in drug delivery and therapeutic efficacy.
KEYWORDS: Area Under Curve, In vitro to in vivo correlation, Maximum plasma concentration, Parallel Artificial Membrane Permeability Assay, Administration Distribution Metabolism and Excretion, Reference Listed Drugs, New Chemical Entity.
INTRODUCTION:
The Protein Data Bank (PDB)1 has become the crucial database that has enabled successful application of machine learning (ML) tools to protein structure prediction. Public availability of scientific data drives research and development.
The artificial intelligence (AI) will continue to benefit from open access to structural, biological, chemical, and biochemical data as new algorithms are applied to predicting small-molecule ligand binding and protein-protein interactions. In addition, there are many databases, such as OpenTarget, Uniprot, PubChem, ChEMBL, DrugBank, SwissTarget, etc. are available to integrate diverse information of molecular pathways, crystal structures, binding affinities, drug targets, disease relevance, chemical properties and biological activities.
A faster and more accurate method in defining disease cure using artificial intelligence and machine deep learning2,3,4 are also under development. All of these efforts in new chemical entity (NCE) development would eventually lead to the dosage forms design, including injection and or oral formulation. Furthermore, the oral formulation is preferred due to its convenience in both carrying and administration. However, the complexity in oral drugs ADME (absorption, distribution, metabolism, and excretion) processes makes drugs formulation difficult to design. Currently, there are various ADME simulation software available to help oral drugs formulation design5,6. Traditionally, dissolution experiments were used to examine the performance of oral drug formulations. However, a number of issues existed, for example, the buffer solutions used in dissolution tests as required per the U.S Food and Drug Administration (FDA) guidance, which often falls short in mimicking conditions in the gastrointestinal (GI) tract such as the various enzymes from bile and pancreatic secretions, resulting in poor in vitro and in vivo correlation (IVIVC) data7,8.
An alternative method that improve IVIVC for oral drugs prediction is Parallel Artificial Membrane Permeability Assay (PAMPA)9, which uses a chemically-based membrane instead of alive cells but has been proven to be able to accurately mimic the human small intestine using biorelevant media10,11,12. A PAMPA Dissolution system (Figure 1) is an instrument that combines dissolution and permeation in a way that closely simulates in vivo conditions. It measures the two necessary parameters—dissolution and permeation—for finding oral drug’s dissolution and permeation. These graphs enable the calculation of Area Under Curve (AUC) and maximum plasma concentration (Cmax) via previously validated equation F (drug absorbed) = Cb*Pe*Area13, which aid in predicting NDA formulation with respect to the RLD (Reference Listed Drugs) that share similar ADME parameters to NCE14,15,16.
A schematic diagram shown in Figure 2 has depicted the proposed flow of a oral NDA formulation development with AI molecular prediction17,18. Details of each step is illustrated in the experimental section.
Figure 1PAMPA Dissolution apparatus
Figure 2 Proposed flow of a oral NDA formulation development
AI Drugs Prediction:
(i) Target Discovery and molecular docking:
Researchers use PDB to explore the spatial arrangements of amino acids, nucleotides, and other molecular components within proteins. By using visualization and analysis of these 3D structures, scientists understand the molecular mechanisms that drive these processes, including enzyme catalysis, protein folding, and molecular recognition. This knowledge helps in developing new pharmaceuticals and the prediction of interactions between ligands and target proteins19,20.
OpenTarget, ChemBL, SwissTarget, SwissADME, Pyrx, ChimeraX are tools in computational biology and drug discovery, each contributing distinct functions to various aspects of the drug development process as shown in Figure 3.
OpenTarget is a platform that integrates genetic and genomic data with molecular pathways that enables researchers to correctly identify potential drug targets implicated in various diseases. Through OpenTarget's approach, researchers gain access to multiple data sets, which analyze disease mechanisms and target identification for drug intervention. Data obtained from ChEMBL can also be seamlessly integrated into OpenTarget, enriching the platform with insights into compound bioactivity and structure-activity relationships. By combining diverse data-sets from ChEMBL and other sources, OpenTarget enhances understanding of complex diseases and helps researchers develop targeted strategies that address medical needs. Smile codes from ChemBL can be used in SwissTarget. The compatibility between SwissTarget and Open Target, facilitated the collaborative nature of biomedical research and drug discovery.
SwissADME is a web tool and application designed to predict pharmacokinetic properties of small molecules. Using SwissADME, researchers can estimate key parameters such as bioavailability, which is critical for determining a compound's properties and potential for further development. By integrating models based on molecular descriptors and algorithms, SwissADME aids researchers in prioritizing lead compounds with optimal pharmacokinetic profiles, accelerating drug discovery while minimizing the risk of in later stages of development.
Pyrx is a software used for virtual screening and molecular docking studies. PyRx provides an interface for performing docking simulations, where researchers can predict the binding affinities and orientations of small molecule ligands with target proteins. By incorporating PyRx's computational capabilities, researchers can explore large chemical libraries and identify potential drug candidates with high binding affinity, expediting the process of lead compound discovery and optimization.
Figure 3 Drug discovery flowchart using AI
ChimeraX is a powerful molecular visualization and analysis tool, particularly renowned for its structural biology and macromolecular modeling applications. Its interface and advanced computational features, Chimerax enables researchers to visualize and manipulate 3D structures of proteins, nucleic acids, and small molecules. Beyond visualization, ChimeraX helps in molecular docking simulations, electrostatic surface calculations, and structural analysis, providing insights into the mechanisms of molecular recognition and interactions. By integrating ChimeraX into drug discovery, researchers can study structure-activity relationships, optimize lead compounds, and design therapeutics with enhanced information.
(ii) SPR and cell lines confirmation:
Using Surface Plasma Resonance (SPR) to confirm the affinity of NCE to the target protein21. Further confirmation of NCE to various cell lines are needed.
(iii) Lead molecule optimization:
In order to estimate oral ADME and Cmax22, especially drugs in the GI track, optimization of NCE molecule23 to meet various enzymes effects should be considered.
Clinical PAD Estimation:
According to the USFDA guidance24, a demonstration of pharmaceutical active dose (PAD) is recommended following MRSD (maximum recommended starting dose). Usually, the ADME simulation with PBPK model is employed to predict PAD for clinical study as shown in Figure 4 (Method (1)). An alternative PAD estimation may also be performed using FDA approved drugs PK database and PAMPA dissolution from J. Chou (Method (2)). In this case, FDA approved PK database is used to compare with predicted ADME data of NCE which lead to the RLD of an approved drug. Therefore, the RLD could be used to perform PAMPA dissolution as the base for NDA formulation development as shown in Figure 4-Method (2).
Figure 4 Prediction of ADME for oral NDA formulation
Conclusion
Integrating AI prediction with target and chemical databases marks a significant shift in new drug discovery, moving away from traditional cell lines and animal preclinical studies. Concurrently, biotech advancements have illuminated the intricate relationship between diseases and their pathways, enhancing our understanding of pathophysiology. Leveraging tools such as AI docking, ADME simulation, PAMPA Dissolution, and the FDA-approved drugs PK database holds promise in optimizing the design of oral NDA drug formulations and improving their PK prediction. These innovative approaches represent a shift in drug development, offering more precise and efficient strategies for addressing challenges.
This work is supported by YQ Biotech, Taiwan and Isuzu Optics, Taiwan.
The authors have no conflicts of interest regarding this investigation.
1. Stephen K. Burley，Helen M. Berman. Open-access data: A cornerstone for artificial intelligence approaches to protein structure prediction. Structure 29, June 3, 2021. https://doi.org/10.1016/j.str.2021.04.010
2. Feng Ren et al. Alpha Fold accelerates artificial intelligence powered drug discovery: efficient discovery ofa novel CDK20 small molecule inhibitor. Chem. Sci. 2023; 14: 1443. DOI: 10.1039/d2sc05709c
3. Elaine C. Meng et al.UCSF ChimeraX: Tools for structure building and analysis. Protein Science. 2023; 32: e4792.https://doi.org/10.1002/pro.4792
4. Neera Borkakoti, Janet M. Thornton. Alphafold2 protein structure prediction : Implications for drug discovery. Curr Opin Struct Biol. 2023 February 06.
5. Michael Davies, Rhys D.O. Jones, Ken Grime, Rasmus Jansson-Löfmark, Adrian J. Fretland, Susanne Winiwarter, Paul Morgan, and Dermot F. McGinnity. Improving the Accuracy of Predicted Human Pharmacokinetics: Lessons Learned from the AstraZeneca Drug Pipeline Over Two Decades. Trends in Pharmacological Sciences. 2020; 41(6). doi:10.1016/j.tips.2020.03.004.
6. Cheng Li, Tongtong Liu, Xiaoming Cui, Annette S. USS, and K.-C. Cheng. Development of In Vitro Pharmacokinetic Screens using Caco-2, Human Hepatocyte, and Caco-2/Human Hepatocyte Hybrid Systems for the Prediction of Oral Bioavailability in Humans. Journal of Biomolecular Screening. 2007: 1084-1091. doi: 10.1177/1087057107308892.
7. Sirisha. S. A Review on IVIVC in The Development of Oral Drug Formulation: Data Obtained from Past Two Decades. Res. J. Pharma. Dosage Forms and Tech. 2020; 12(3): 198-204.
8. AV Bhosale, SR Hardikar, Naresh Patil, Rajesh Jagtap, Nilesh Jamdade, Bhavin Patel. IVIVC and BCS: A Regulatory Perspective. Research J. Pharm. and Tech. 2009; 2(1): 72-79.
9. Sugano K, Hamada H, Machida M, Ushio H, Saitoh K, Terada K. Optimized conditions of bio-mimetic artificial membrane permeation assay. Int J Pharm. 2001 Oct 9; 228(1-2): 181-8. doi: 10.1016/s0378-5173(01)00845-6. PMID: 11576780.
10. Jantratid E, Janssen N, Reppas C, Dressman JB. Dissolution media simulating conditions in the proximal human gastrointestinal tract: an update. Pharm Res. 2008 Jul; 25(7): 1663-76. doi: 10.1007/s11095-008-9569-4. Epub 2008 Apr 11. PMID: 18404251.
11. Manju Nagpal, Pankaj Rakha, Surinder Goyal, Gitika Dhingra, Sunil Gupta. Comparison of Biorelevant and Compendial Dissolution Media and Prediction of In-vivo Plasma Profile of BCS Class II Drug. Research J. Pharma. Dosage Forms and Tech. 2010; 2(1): 37-40 .
12. Gitika Dhingra, Brinda Sreelesh, Manju Nagpal, Pankaj Rakha, B P Nagori. In-Vitro Dissolution Testing of Ibuprofen Using Compendial and Biorelevant Dissolution Media. Research J. Pharm. and Tech. 2010; 3(3): 931-933.
13. Sinko, Patrick & Leesman, Glen and Amidon, Gordon. (1991). Predicting Fraction Dose Absorbed in Humans Using a Macroscopic Mass Balance Approach. Pharmaceutical Research. 8. Doi：10.1023/A:1015892621261.
14. Joe Chou et al. PAMPA Dissolution: An alternative method for oral NDA formulation development. RJPDFT. 2024; 16(2).
15. Saptarshi Samajdar. In silico ADME prediction of Phytochemicals present in Piper longum. Asian Journal of Research in Pharmaceutical Sciences. 2023; 13(2):89-3.
16. Pooja Sharma, Navneet Bithel. Phytochemical screening, Antimicrobial potential, ADME predictions Targeting the Helicteres isora L. fruit extracts and their Essential oil. Research Journal of Pharmacy and Technology. 2023; 16(11): 5015-0.
17. Vishwajit S. Patil, Prithviraj A. Patil. Molecular Docking: A useful approach of Drug Discovery on the Basis of their Structure. Asian Journal of Pharmaceutical Research. 2023; 13(3):191-5.
18. Soham Pawar, Chaitrali Kulkarni, Puja Gadade, Supriya Pujari, Surajkumar Kakade, S.H. Rohane, V.K. Redasani. Molecular Docking using different Tools. Asian Journal of Pharmaceutical Research. 2023; 13(4): 292-6.
19. Rutuja P. Pawar, Sachin H. Rohane. Role of Autodock vina in PyRx Molecular Docking. Asian J. Research Chem. 2021; 14(2): 132-134.
20. Shravani S. Pawar, Sachin H. Rohane. Review on Discovery Studio: An important Tool for Molecular Docking. Asian J. Research Chem. 2021; 14(1): 86-88.
21. Senem Aykul and Erik Martinez-Hackert. Determination of half-maximal inhibitory concentration using biosensor-based protein interaction analysis. Anal Biochem. 2016 September 1; 508: 97–103. doi:10.1016/j.ab.2016.06.025.
22. Ryan M. Fryer. Mathematical Derivation of Therapeutic Index Based on hERG IC50 Taking Into Account Fraction Unbound in Plasma and Clinical Cmax. Drug Information Journal. 46(5). DOI: 10.1177/0092861512456285
23. Xiao-han Song; Jiang Wang; Hong Liu. Lead compound optimization strategy (8) - drug transporters and related drug design strategies. Acta Pharmaceutica Sinica. 2021; 12: 432-444.
24. Guidance for Industry: Estimating the Maximum Safe Starting Dose in Initial Clinical Trials for Therapeutics in Adult Healthy Volunteers. Center for Drug Evaluation and Research (CDER), USFDA, 2005.
Received on 16.03.2024 Modified on 20.05.2024
Accepted on 04.07.2024 ©AandV Publications All Right Reserved
Res. J. Pharma. Dosage Forms and Tech.2024; 16(3):289-292.
DOI: 10.52711/0975-4377.2024.00045