Development of Co-Amorphous Systems for Inhalation Therapy – Part 1: From Model Prediction to Clinical Success

Abstract

Background/Objectives: The integration of machine learning (ML) and artificial intelligence (AI) has revolutionized the pharmaceutical industry by improving drug discovery, development and manufacturing processes. Based on literature data, an ML model was developed by our group to predict the formation of binary co-amorphous systems (COAMSs) for inhalation therapy. The model’s ability to develop a dry powder formulation with the necessary properties for a predicted co-amorphous combination was evaluated.

Methods: An extended experimental validation of the ML model by co-milling and X-ray diffraction analysis for 18 API-API (active pharmaceutical ingredient) combinations is presented. Additionally, one COAMS of rifampicin (RIF) and ethambutol (ETH), two first-line tuberculosis (TB) drugs are developed further for inhalation therapy.

Results: The ML model has shown an accuracy of 79% in predicting suitable combinations for 35 APIs used in inhalation therapy; experimental accuracy was demonstrated to be 72%. The study confirmed the successful development of stable COAMSs of RIF-ETH either via spray-drying or co-milling. In particular, the milled COAMSs showed better aerosolization properties (higher ED and FPF with lower standard deviation). Further, RIF-ETH COAMSs show much more reproducible results in terms of drug quantity dissolved over time.

Conclusions: ML has been shown to be a suitable tool to predict COAMSs that can be developed for TB treatment by inhalation to save time and cost during the experimental screening phase.

Introduction

The integration of Machine Learning (ML) and Artificial Intelligence (AI) into the pharmaceutical industry has been transformative, enabling breakthroughs across various phases of drug discovery, development, and manufacturing [1]. These advanced technologies leverage vast amounts of biological, chemical, and clinical data to make predictions, identify patterns, and optimize processes that would be virtually impossible to achieve using traditional methods [2]. The convergence of AI and ML has revolutionized the industry’s approach to drug discovery and design [3,4], clinical trials [5], personalized medicine [6], and even regulatory processes [7]. In the past, the pharmaceutical industry relied heavily on experimental trial-and-error methods to uncover potential drug candidates and validate their efficacy. However, with the rise of ML and AI, predictive models are now used to simulate biological systems, forecast drug interactions, and estimate the probability of clinical success. This shift from purely empirical experiments to data-driven predictions is arguably one of the most significant tends in pharmaceutical innovation today. AI-powered platforms are now able to quickly sift through vast chemical libraries, identify promising drug candidates and predict potential adverse effects even before the physical testing phase. Artificial neural networks (e.g., Chem software) are used in controlled-release tablet formulation, hard-capsule shell formation, solid dispersions, pellets, and micro- and nanoparticles [8].

The collaboration between advanced algorithms and experimental validation remains essential to ensure that AI- and ML-driven predictions translate effectively into real-world therapeutic solutions. So far, current AI- or ML-based approaches are not a substitute for traditional experimental methods; however, AI and ML can provide predictions based on the data available and potentially limit and facilitate experimental efforts, as the predicted outcome must then be validated and interpreted by human researchers [9]. The integration of AI and ML with traditional experimental methods can enhance pharmaceutical processes like drug discovery processes [4,10], accelerate the development of new medications [3,11], and optimize and control pharmaceutical unit operations during manufacturing [12,13]. Especially for pulmonary drug delivery systems, where the testing in laboratory animals is less predictive than for other administration routes, in silico tools may be very useful [14]. Treatments with more than one drug are common in the treatment of respiratory diseases and combined administration will increase patient compliance. However, the generation of such formulations may pose problems.

For this reason, we have developed a simple predictive ML tool for the formation of co-amorphous systems (COAMSs) to support scientists and enable rapid screening and minimize laboratory effort, time and cost [15]. A molecular descriptor-based ML model for predicting the potential of binary drug combinations to form co-amorphous systems was built based on available literature data (generation of a literature database). In contrast to previously reported predictive models, input data from four classes of COAMS (active pharmaceutical ingredient (API)-API (1), API—amino acid (2) API—organic acid (3) and API—other substance (4)), making the model relatively universally applicable. The accuracy of the generated ML model was 79%. Predictions were made for 35 active pharmaceutical ingredients (APIs) used in inhalation therapy [15]. The inhalation route was taken as in typical lung diseases, namely asthma, chronic obstructive pulmonary diseases (COPDs) and tuberculosis (TB), it is common to apply multiple drugs and combinations over a longer period [16,17,18,19]. COAMSs for inhalation therapy are suggested to improve patient compliance by reducing the number of medications to take and decrease the drug dose variability while administering combination products to patients. Further, COAMSs have advantageous properties, such as improved solubility, dissolution, and stability [20]. They are defined as homogeneous single-phase systems where typically an API is combined with a co-former (low-molecular-weight compound) and the system is stabilized. A co-former can either be another API or an excipient (e.g., amino acid, organic acid). The selection of suitable co-formers is crucial for successful co-formability and a lack of systematic, predictive, and computational methods for this selection has been identified [21]. An overview on predictive models available for co-former selection can be found in our previous paper [15].

So far only three systems have been tested experimentally and published together with model building [15]. In the present work, a further 15 systems were selected for in-depth model validation. Further, one relevant therapeutic system was selected and developed as a dry powder for inhalation therapy. To demonstrate the practical value of the model, a combination of two first-line drugs for TB treatment, rifampicin (RIF) and ethambutol (ETH), was selected due to the global importance of TB and the lack of inhaled formulations for TB treatment. TB is one of the major global health burdens, caused by Mycobacterium tuberculosis and mainly affecting the lungs, but it can also spread to other parts of the body [22]. Before the COVID-19 pandemic, TB was the most prevalent infectious disease worldwide [23]. To date, the standard therapy is still oral or parenteral, involving various antibiotics over a long period of time. However, there is a big research focus on TB treatment via inhalation [24,25,26,27]. Delivering TB drugs via inhalation could lower systemic doses while achieving higher lung doses (site of action and primary location of disease) and efficacy [28]. Although tuberculostatic drugs for inhalation were tested in preclinical studies, few formulations, e.g., dry powder amikacin (NCT04249531) and nebulized rifampicin (NCT06041919), have been evaluated in clinical trials, and no formulation has reached the market yet.

Our goal is to develop an innovative drug–drug COAMS for inhalation therapy of TB. Therefore, an extended validation of the predictive ML tool to predict COAMS (developed by our group and published previously in Pharmaceutics) was first conducted. Based on the results, a promising co-amorphous combination of RIF-ETH was selected further and developed as powder for inhalation therapy. For efficient delivery to the lung, particles should be 1–5 µm [29]. Therefore, particle size, solid state, morphology, aerosolization performance, and dissolution of the developed dry powder formation were determined.

Download the full article as PDF here: Development of Co-Amorphous Systems for Inhalation Therapy

or read it here

Materials

The following active ingredients were used for model validation: salbutamol sulfate (SBS) (Fagron GmbH, Glinde, Germany), glycopyrronium bromide (GB) (kindly donated from Chiesi Pharmaceuticals, Parma, Italy), mometasone fuorate (MOM), bambuterol HCl (BAM) (both from Shenzhen Nexconn Pharmatechs Ltd., Shenzhen, China), isoniazid (ISO), streptomycin sulfate (STR), ethambutol (ETH), pyrazinamide (PYR), budesonide (all purchased from TCI Deutschland GmbH, Eschborn, Germany) and rifampicin (RIF) (Olon Active Pharmaceutical Ingredients, Mumbai, India).

Inhalac 500 (Meggle GmbH, Wasserburg, Germany), Magnesium Stearate ((MgSt) (Sigma-Aldrich, Darmstadt, Germany)) and Leucin (Merck KGaA, Darmstadt, Germany) were used as fine excipients.

Solvents were used for analytics (acetic acid, acetonitrile, methanol) and spray-drying (absolute Ethanol). Absolute ethanol was purchased from Lactan Chemikalien and Laborgeraete Vertriebsgesellschaft m.b.H & Co. KG (Graz, Austria).

For the preparation of simulated lung fluid (SLF), Gibco™ PBS, pH 7.4 buffer (Fisher Scientific GmbH, Vienna, Austria) and 1,2-Dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) (TCI Deutschland GmbH, Eschborn, Germany) were used.

For aerosolization tests hard-gelatin Coni-Snap^® size 3 capsules provided by Capsugel (Köln, Germany) and a capsule-based inhaler, the Cyclohaler^® (PB Pharma GmbH, Meerbusch, Germany), were used.

Source: Fröhlich, E.; Bordoni, A.; Mohsenzada, N.; Mitsche, S.; Schröttner, H.; Zellnitz-Neugebauer, S. Development of Co-Amorphous Systems for Inhalation Therapy—Part 1: From Model Prediction to Clinical Success. Pharmaceutics 2025, 17, 922. https://doi.org/10.3390/pharmaceutics17070922