Predicting drug solubility in organic solvents mixtures: A machine-learning approach supported by high-throughput experimentation.

A novel approach based on supervised machine-learning is proposed to predict the solubility of drugs and drug-like molecules in mixtures of organic solvents. Similar to quantitative structure-property relationship (QSPR) models, different solvent types are identified by molecular descriptors, which, in this study, are considered as UNIFAC subgroups. To overcome the potential lack of UNIFAC subgroups for the complex Active Pharmaceutical Ingredients (APIs) currently developed in the pharmaceutical industry, the API molecule is considered as a unique entity in the proposed modelling approach. Therefore, API solubility is predicted as a function of temperature, functional subgroups of the solvents and composition of the solvent mixture; in turn, regressors' correlation is handled through Partial Least-Squares (PLS) regression. The method is developed and tested with experimental data of a real API and 14 organic solvents that are industrially employed for crystallisation. Solubility predictions are accurate and precise for single solvents, binary mixtures and ternary mixtures of organic solvents at different compositions and temperatures, with a determination coefficient R2 ≥ 0.90. To further test the applicability of the model, the proposed approach is applied to 9 literature organic solubility datasets of drugs and drug-like compounds and compared to benchmark solubility models in the literature. Results show that the proposed approach provides satisfactory predictions: the majority of validation and calibration data have R2 = 0.95-0.99; the ratio between RMSE (root mean squared error) of the proposed method and the range of measured solubility values is from 1 to 3 orders of magnitude smaller than the RMSE ratio obtained by the benchmark models.

Quality-by-control of intensified continuous filtration-drying of active pharmaceutical ingredients.

Continuous manufacturing and closed-loop quality control are emerging technologies that are pivotal for next-generation pharmaceutical modernization. We develop a process control framework for a continuous carousel for integrated filtration-drying of crystallization slurries. The proposed control system includes model-based monitoring and control routines, such as state estimation and real-time optimization, implemented in a hierarchical, three-layer quality-by-control (QbC) framework. We implement the control system in ContCarSim, a publicly available carousel simulator. We benchmark the proposed control system against simpler methods, comprising a reduced subset of the elements of the overall control system, and against open-loop operation (the current standard in pharmaceutical manufacturing). The proposed control system demonstrates superior performance in terms of higher consistency in product quality and increased productivity, proving the benefits of closed-loop control and of model-based techniques in pharmaceutical manufacturing. This study represents a step forward toward end-to-end continuous pharmaceutical processing, and in the evolution of quality-by-design toward quality-by-control.

A review on the modernization of pharmaceutical development and manufacturing - Trends, perspectives, and the role of mathematical modeling.

Recently, the pharmaceutical industry has been facing several challenges associated to the use of outdated development and manufacturing technologies. The return on investment on research and development has been shrinking, and, at the same time, an alarming number of shortages and recalls for quality concerns has been registered. The pharmaceutical industry has been responding to these issues through a technological modernization of development and manufacturing, under the support of initiatives and activities such as quality-by-design (QbD), process analytical technology, and pharmaceutical emerging technology. In this review, we analyze this modernization trend, with emphasis on the role that mathematical modeling plays within it. We begin by outlining the main socio-economic trends of the pharmaceutical industry, and by highlighting the life-cycle stages of a pharmaceutical product in which technological modernization can help both achieve consistently high product quality and increase return on investment. Then, we review the historical evolution of the pharmaceutical regulatory framework, and we discuss the current state of implementation and future trends of QbD. The pharmaceutical emerging technology is reviewed afterwards, and a discussion on the evolution of QbD into the more effective quality-by-control (QbC) paradigm is presented. Further, we illustrate how mathematical modeling can support the implementation of QbD and QbC across all stages of the pharmaceutical life-cycle. In this respect, we review academic and industrial applications demonstrating the impact of mathematical modeling on three key activities within pharmaceutical development and manufacturing, namely design space description, process monitoring, and active process control. Finally, we discuss some future research opportunities on the use of mathematical modeling in industrial pharmaceutical environments.

Integrating metabolome dynamics and process data to guide cell line selection in biopharmaceutical process development.

The successful development of mammalian cell culture for the production of therapeutic antibodies is a resource-intensive and multistage process which requires the selection of high performing and stable cell lines at different scale-up stages. Accordingly, science-based approaches exploiting biological information, such as metabolomics, can support and accelerate the selection of promising cell lines to progress. In fact, the integration of dynamic biological information with process data can provide valuable insights on the cell physiological changes as a consequence of the cultivation process. This work studies the industrial development of monoclonal antibodies at micro-bioreactor scale (Ambr®15) and aims at accelerating the selection of the better performing cell lines. To that end, we apply a machine learning approach to integrate time-varying process and biological information (i.e., metabolomics), explicitly exploiting their dynamics. Strikingly, cell line performance during the cultivation can be predicted from early process timepoints by exploiting the gradual temporal evolution of metabolic phenotypes. Furthermore, product titer is estimated with good accuracy at late process timepoints, providing insights into its relationship with underlying metabolic mechanisms and enabling the identification of biomarkers to be further investigated. The biological insights obtained through the proposed machine learning approach provide data-driven metabolic understanding allowing early identification of high performing cell lines. Additionally, this analysis offers the opportunity to identify key metabolites which could be used as biomarkers for industrially relevant phenotypes and onward fit into our commercial manufacturing platforms.

Practical use of primary drying models in an industrial environment with limited availability of equipment sensors.

In the pharmaceutical industry, lyophilization is typically adopted to extend long-time stability of valuable thermolabile medicines and vaccines. Primary drying is the most time-consuming and energy-intensive step of the entire process; thus, accelerating and optimizing the primary drying recipe is a key process development goal. To that purpose, mathematical models have been proposed and successfully validated. However, models typically require invasive experiments and/or sensors (e.g. product temperatures) for parameter estimation, which are rarely available in good manufacturing practice (GMP) environment. This represents a severe limitation when leveraging the model to transfer operation recipes across different facilities and for scale-up. In this study, we assess the possibility to exploit limited industrial data for model parameter estimation, namely pressure measurements and gravimetric tests, by defining a calibration protocol that is tested on two different pieces of equipment. Results are verified on a recently proposed model, and show that statistically meaningful estimates can be obtained without the need of product temperature measurements. Model predictions and optimal inputs trajectories are comparable to those obtained from the model calibrated using the full set of temperature and pressure data.

Streamlining tablet lubrication design via model-based design of experiments.

In oral solid dosage production through direct compression powder lubrication must be carefully selected to facilitate the manufacturing of tablets without degrading product manufacturability and quality (e.g. dissolution). To do so, several semi-empirical models relating compression performance to process operating conditions have been developed. Among them, we consider an extension of the Kushner and Moore model (Kushner and Moore, 2010, International Journal Pharmaceutics, 399:19) that is useful for the purpose, but requires an extensive experimental campaign for parameters identification. This implies the preparation and compression of multiple powder blends, each one with a different lubrication extent. In turn, this translates into a considerable consumption of Active Pharmaceutical Ingredient (API), and into time-consuming experiments. We tackled this issue by proposing a novel model-based design of experiments (MBDoE) approach, which minimizes the number of optimal blends for model calibration, while obtaining statistically sound parameters estimates and model predictions. Both sequential and parallel MBDoE configurations were compared. Experimental results involving two placebo blends with different lubrication sensitivity showed that this methodology is able to reduce the experimental effort by 60-70% with respect to the standard industrial practice independently of the formulation considered and configuration (i.e. parallel vs. sequential) adopted.

A benchmark simulator for quality-by-design and quality-by-control studies in continuous pharmaceutical manufacturing - Intensified filtration-drying of crystallization slurries.

This article introduces ContCarSim, a benchmark simulator for the development and testing of quality-by-design and quality-by-control strategies in the continuous intensified filtration-drying of paracetamol/ethanol slurries on a novel carousel technology, developed by Alconbury Weston Ltd (United Kingdom). The simulator is based on a detailed mechanistic mathematical modeling framework, and has been validated with filtration and drying experiments on a prototype equipment. A set of design- and control-relevant challenges to be addressed through ContCarSim are proposed. A case study is developed, to demonstrate the features of the simulator and its suitability to design, test and optimize the unit operation. ContCarSim is expected to promote the transition to end-to-end continuous pharmaceutical manufacturing and the adoption of closed-loop quality control by the pharmaceutical industry. The simulator can also be employed as a benchmark for data analytics and process monitoring studies.

Optimization of the Appearance Quality in CO2 Processed Ready-to-Eat Carrots through Image Analysis.

A high-pressure CO2 process applied to ready-to-eat food products guarantees an increase of both their microbial safety and shelf-life. However, the treatment often produces unwanted changes in the visual appearance of products depending on the adopted process conditions. Accordingly, the alteration of the visual appearance influences consumers' perception and acceptability. This study aims at identifying the optimal treatment conditions in terms of visual appearance by using an artificial vision system. The developed methodology was applied to fresh-cut carrots (Daucus carota) as the test product. The results showed that carrots packaged in 100% CO2 and subsequently treated at 6 MPa and 40 °C for 15 min maintained an appearance similar to the fresh product for up to 7 days of storage at 4 °C. Mild appearance changes were identified at 7 and 14 days of storage in the processed products. Microbiological analysis performed on the optimal treatment condition showed the microbiological stability of the samples up to 14 days of storage at 4 °C. The artificial vision system, successfully applied to the CO2 pasteurization process, can easily be applied to any food process involving changes in the appearance of any food product.

Powder composition monitoring in continuous pharmaceutical solid-dosage form manufacturing using state estimation - Proof of concept.

In continuous solid-dosage form manufacturing, the powder feeding system is responsible for supplying downstream the correct formulation of the drug product ingredients. The composition of the powder delivered by the feeding system is inferred from the measurements of powder mass flow from the system feeders. The mass flows are, in turn, inferred from the loss in weight measured in the feeder hoppers. Most loss-in-weight feeders post-process the mass flow signal to deliver a smoothed value to the user. However, such estimated mass flows can exhibit a low signal-to-noise ratio. As the feeders are critical elements of the control strategy of the manufacturing line, better instantaneous estimates of mass flow are desirable for improving the quality assurance. In this study, we propose a model-based approach for monitoring the composition of the powder fed to a continuous solid-dosage line. The monitoring system is based on a moving-horizon state estimator, which carries out model-based reconciliation of the feeder mass measurements, thus enabling accurate composition estimation of the powder mixture. Experimental datasets from a direct compression line are used to validate the methodology. Results demonstrate improvement with respect to current industrial solutions.

Mathematical modeling and digital design of an intensified filtration-washing-drying unit for pharmaceutical continuous manufacturing.

This paper introduces a comprehensive mathematical model of a novel integrated filter-dryer carousel system, designed for continuously filtering, washing and drying a slurry stream into a crystals cake. The digital twin includes models for dead-end filtration, cake washing and convective cake drying, based on dynamic multi-component mass, energy and momentum balances. For set of feed conditions and control inputs, the model allows tracking the solvents and impurities content in the cake (critical quality attributes, CQAs) throughout the whole process. The model parameters were identified for the isolation of paracetamol from a multi-component slurry, containing a non-volatile impurity. The calibrated model was used for identifying the probabilistic design space and maximum throughput for the process, expressing the combinations of the carousel feed conditions and control inputs for which the probability of meeting the target CQAs is acceptable.

Data analytics on raw material properties to accelerate pharmaceutical drug development.

Manufacturability of active pharmaceutical ingredients (APIs) is often evaluated by an empirical approach during development due to limited material availability. This brings challenges in designing flexible yet robust manufacturing processes under highly accelerated timelines. Hence, good utilisation of a limited material dataset is key to accelerate the delivery of high quality final drug product into the market at minimum cost and maximum process capacity. In this study, we present a data-driven method to investigate a raw materials database where the integration of multivariate analysis and machine learning modelling aids the selection of new incoming materials based on their manufacturability. The procedure was applied to an industrial representative database of thirty-four APIs and seven excipients where eight measurements relevant to flow properties for each of those forty-one materials were collected. The models identified four clusters of materials with different flow properties. These models can serve as a risk assessment tool for new API in early product development phases based on the nearest surrogate material which behave similarly, as well as to identify targeted and material sparring experiments to address key risks during secondary process selection.

A Mechanistic Model to Quantify von Willebrand Factor Release, Survival and Proteolysis in Patients with von Willebrand Disease.

A reduced von Willebrand factor (VWF) synthesis or survival, or its increased proteolysis, alone or in combination, contributes to the development of von Willebrand disease (VWD).We describe a new, simple mechanistic model for exploring how VWF behaves in well-defined forms of VWD after its 1-desamino-8-D-arginine vasopressin (DDAVP)-induced release from endothelial cells. We aimed to ascertain whether the model can consistently predict VWF kinetic changes. The study involved 9 patients with VWD types Vicenza (a paradigmatic form with a reduced VWF survival), 8 type 2B, 2 type 2A-I, 1 type 2A-II (associated with an increased VWF proteolysis), and 42 normal controls, whose VWF levels were measured after a 24-hour-long DDAVP test. The rate constants considered were: k0, associated with the VWF release phase; k1, illustrating the phase of conversion from high- to low-molecular-weight VWF multimers; and ke, associated with the VWF elimination phase. The amount of VWF released (D) was also measured. ke and D were significantly higher in O than in non-O blood group controls; k1 was also higher, but less markedly so. All the parameters were accelerated in type Vicenza, especially ke (p < 0.0001), which explains the significant reduction in VWF half-life. In types 2B and 2A-II, k1 was one order of magnitude higher than in controls, which explains their loss of large VWF multimers. All parameters except ke were lower in type 2A-I.The proposed mechanistic model clearly describes the altered biochemical pathways in well-characterized VWD, prompting us to suggest that it might help clarify elusive forms of VWD too.

Assessing the relative potency of (S)- and (R)-warfarin with a new PK-PD model, in relation to VKORC1 genotypes.

PURPOSE: The purpose of this study is to develop a new pharmacokinetic-pharmacodynamic (PK-PD) model to characterise the contribution of (S)- and (R)-warfarin to the anticoagulant effect on patients in treatment with rac-warfarin. METHODS: Fifty-seven patients starting warfarin (W) therapy were studied, from the first dose and during chronic treatment at INR stabilization. Plasma concentrations of (S)- and (R)-W and INRs were measured 12, 36 and 60 h after the first dose and at steady state 12-14 h after dosing. Patients were also genotyped for the G>A VKORC1 polymorphism. The PK-PD model assumed a linear relationship between W enantiomer concentration and INR and included a scaling factor k to account for a different potency of (R)-W. Two parallel compartment chains with different transit times (MTT1 and MTT2) were used to model the delay in the W effect. PD parameters were estimated with the maximum likelihood approach. RESULTS: The model satisfactorily described the mean time-course of INR, both after the initial dose and during long-term treatment. (R)-W contributed to the rac-W anticoagulant effect with a potency of about 27% that of (S)-W. This effect was independent of VKORC1 genotype. As expected, the slope of the PK/PD linear correlation increased stepwise from GG to GA and from GA to AA VKORC1 genotype (0.71, 0.90 and 1.49, respectively). CONCLUSIONS: Our PK-PD linear model can quantify the partial pharmacodynamic activity of (R)-W in patients contemporaneously exposed to therapeutic (S)-W plasma levels. This concept may be useful in improving the performance of future algorithms aiming at identifying the most appropriate W maintenance dose.

Knowledge management in secondary pharmaceutical manufacturing by mining of data historians-A proof-of-concept study.

In this proof-of-concept study, a methodology is proposed to systematically analyze large data historians of secondary pharmaceutical manufacturing systems using data mining techniques. The objective is to develop an approach enabling to automatically retrieve operation-relevant information that can assist the management in the periodic review of a manufactory system. The proposed methodology allows one to automatically perform three tasks: the identification of single batches within the entire data-sequence of the historical dataset, the identification of distinct operating phases within each batch, and the characterization of a batch with respect to an assigned multivariate set of operating characteristics. The approach is tested on a six-month dataset of a commercial-scale granulation/drying system, where several millions of data entries are recorded. The quality of results and the generality of the approach indicate that there is a strong potential for extending the method to even larger historical datasets and to different operations, thus making it an advanced PAT tool that can assist the implementation of continual improvement paradigms within a quality-by-design framework.

Quality-by-Design approach to monitor the operation of a batch bioreactor in an industrial avian vaccine manufacturing process.

Monitoring batch bioreactors is a complex task, due to the fact that several sources of variability can affect a running batch and impact on the final product quality. Additionally, the product quality itself may not be measurable on line, but requires sampling and lab analysis taking several days to be completed. In this study we show that, by using appropriate process analytical technology tools, the operation of an industrial batch bioreactor used in avian vaccine manufacturing can be effectively monitored as the batch progresses. Multivariate statistical models are built from historical databases of batches already completed, and they are used to enable the real time identification of the variability sources, to reliably predict the final product quality, and to improve process understanding, paving the way to a reduction of final product rejections, as well as to a reduction of the product cycle time. It is also shown that the product quality "builds up" mainly during the first half of a batch, suggesting on the one side that reducing the variability during this period is crucial, and on the other side that the batch length can possibly be shortened. Overall, the study demonstrates that, by using a Quality-by-Design approach centered on the appropriate use of mathematical modeling, quality can indeed be built "by design" into the final product, whereas the role of end-point product testing can progressively reduce its importance in product manufacturing.

Latent variable modeling to assist the implementation of Quality-by-Design paradigms in pharmaceutical development and manufacturing: a review.

The introduction of the Quality-by-Design (QbD) initiative and of the Process Analytical Technology (PAT) framework by the Food and Drug Administration has opened the route to the use of systematic and science-based approaches to support pharmaceutical development and manufacturing activities. In this review we discuss the role that latent variable models (LVMs) can play in the practical implementation of QbD paradigms in the pharmaceutical industry, and the potential they may have in assisting the development and manufacturing of new products. The ultimate scope is to provide practitioners with a perspective on the effectiveness of the use of LVMs in any phase of the development of a pharmaceutical product, from its design up to its commercial production. After an overview of the main regulatory paradigms the QbD initiative is founded on, we show how LVMs can be feasibly used to support pharmaceutical development and manufacturing activities while matching the regulatory Agencies' requirements. Three main areas are identified, wherein the use of LVMs can provide significant benefits: (i) process understanding, (ii) product and process design, and (iii) process monitoring and control. For each of them, the main contributions recently appeared in the literature are reviewed. Issues open for further research are also identified.

A correlative study on data from pork carcass and processed meat (Bauernspeck) for automatic estimation of chemical parameters by means of near-infrared spectroscopy.

The use of near-infrared spectroscopy (NIRS) is proposed in this study for the characterization of the quality parameters of a smoked and dry-cured meat product known as Bauernspeck (originally from Northern Italy), as well as of some technological traits of the pork carcass used for its manufacturing. In particular, NIRS is shown to successfully estimate several key quality parameters (including water activity, moisture, dry matter, ash and protein content), suggesting its suitability for real time application in replacement of expensive and time consuming chemical analysis. Furthermore, a correlative approach based on canonical correlation analysis was used to investigate the spectral regions that are mostly correlated to the characteristics of interest. The identification of these regions, which can be linked to the absorbance of the main functional chemical groups, is intended to provide a better understanding of the chemical structure of the substrate under investigation.

A general model-based design of experiments approach to achieve practical identifiability of pharmacokinetic and pharmacodynamic models.

The use of pharmacokinetic (PK) and pharmacodynamic (PD) models is a common and widespread practice in the preliminary stages of drug development. However, PK-PD models may be affected by structural identifiability issues intrinsically related to their mathematical formulation. A preliminary structural identifiability analysis is usually carried out to check if the set of model parameters can be uniquely determined from experimental observations under the ideal assumptions of noise-free data and no model uncertainty. However, even for structurally identifiable models, real-life experimental conditions and model uncertainty may strongly affect the practical possibility to estimate the model parameters in a statistically sound way. A systematic procedure coupling the numerical assessment of structural identifiability with advanced model-based design of experiments formulations is presented in this paper. The objective is to propose a general approach to design experiments in an optimal way, detecting a proper set of experimental settings that ensure the practical identifiability of PK-PD models. Two simulated case studies based on in vitro bacterial growth and killing models are presented to demonstrate the applicability and generality of the methodology to tackle model identifiability issues effectively, through the design of feasible and highly informative experiments.

General procedure to aid the development of continuous pharmaceutical processes using multivariate statistical modeling - an industrial case study.

Streamlining the manufacturing process has been recognized as a key issue to reduce production costs and improve safety in pharmaceutical manufacturing. Although data available from earlier developmental stages are often sparse and unstructured, they can be very useful to improve the understanding about the process under development. In this paper, a general procedure is proposed for the application of latent variable statistical methods to support the development of new continuous processes in the presence of limited experimental data. The proposed procedure is tested on an industrial case study concerning the development of a continuous line for the manufacturing of paracetamol tablets. The main driving forces acting on the process are identified and ranked according to their importance in explaining the variability in the available data. This improves the understanding about the process by elucidating how different active pharmaceutical ingredient pretreatments, different formulation modes and different settings on the processing units affect the overall operation as well as the properties of the intermediate and final products. The results can be used as a starting point to perform a comprehensive and science-based quality risk assessment that help to define a robust control strategy, possibly enhanced with the integration of a design space for the continuous process at a later stage.

On the use of continuous glucose monitoring systems to design optimal clinical tests for the identification of type 1 diabetes models.

The identification of individual parameters of detailed physiological models of type 1 diabetes can be carried out by clinical tests designed optimally through model-based design of experiments (MBDoE) techniques. So far, MBDoE for diabetes models has been considered for discrete glucose measurement systems only. However, recent advances on sensor technology allowed for the development of continuous glucose monitoring systems (CGMSs), where glucose measurements can be collected with a frequency that is practically equivalent to continuous sampling. To specifically address the features of CGMSs, in this paper the optimal clinical test design problem is formulated and solved through a continuous, rather than discrete, approach. A simulated case study is used to assess the impact of CGMSs both in the optimal clinical test design problem and in the subsequent parameter estimation for the identification of a complex physiological model of glucose homeostasis. The results suggest that, although the optimal design of a clinical test is simpler if continuous glucose measurements are made available through a CGMS, the noise level and formulation may make continuous measurements less suitable for model identification than their discrete counterparts.