Artificial Intelligence: From Buzzword to Useful Tool in Clinical Pharmacology.

The advent of artificial intelligence (AI) in clinical pharmacology and drug development is akin to the dawning of a new era. Previously dismissed as merely technological hype, these approaches have emerged as promising tools in different domains, including health care, demonstrating their potential to empower clinical pharmacology decision making, revolutionize the drug development landscape, and advance patient care. Although challenges remain, the remarkable progress already made signals that the leap from hype to reality is well underway, and AI promises to offer clinical pharmacology new tools and possibilities for optimizing patient care is gradually coming to fruition. This review dives into the burgeoning world of AI and machine learning (ML), showcasing different applications of AI in clinical pharmacology and the impact of successful AI/ML implementation on drug development and/or regulatory decisions. This review also highlights recommendations for areas of opportunity in clinical pharmacology, including data analysis (e.g., handling large data sets, screening to identify important covariates, and optimizing patient population) and efficiencies (e.g., automation, translation, literature curation, and training). Realizing the benefits of AI in drug development and understanding its value will lead to the successful integration of AI tools in our clinical pharmacology and pharmacometrics armamentarium.

Qualifying a Novel Clinical Trial Endpoint (iBOX) Predictive of Long-Term Kidney Transplant Outcomes.

New immunosuppressive therapies that improve long-term graft survival are needed in kidney transplant. Critical Path Institute's Transplant Therapeutics Consortium received a qualification opinion for the iBOX Scoring System as a novel secondary efficacy endpoint for kidney transplant clinical trials through European Medicines Agency's qualification of novel methodologies for drug development. This is the first qualified endpoint for any transplant indication and is now available for use in kidney transplant clinical trials. Although the current efficacy failure endpoint has typically shown the noninferiority of therapeutic regimens, the iBOX Scoring System can be used to demonstrate the superiority of a new immunosuppressive therapy compared to the standard of care from 6 months to 24 months posttransplant in pivotal or exploratory drug therapeutic studies.

Qualifying a novel clinical trial endpoint (iBOX) predictive of long-term kidney transplant outcomes.

New immunosuppressive therapies that improve long-term graft survival are needed in kidney transplant. Critical Path Institute's Transplant Therapeutics Consortium received a qualification opinion for the iBOX Scoring System as a novel secondary efficacy endpoint for kidney transplant clinical trials through European Medicines Agency's qualification of novel methodologies for drug development. This is the first qualified endpoint for any transplant indication and is now available for use in kidney transplant clinical trials. Although the current efficacy failure endpoint has typically shown the noninferiority of therapeutic regimens, the iBOX Scoring System can be used to demonstrate the superiority of a new immunosuppressive therapy compared to the standard of care from 6 months to 24 months posttransplant in pivotal or exploratory drug therapeutic studies.

T1dCteGui: A User-Friendly Clinical Trial Enrichment Tool to Optimize T1D Prevention Studies by Leveraging AI/ML Based Synthetic Patient Population.

Whereas islet autoantibodies (AAs) are well-established risk factors for developing type 1 diabetes (T1D), there is a lack of biomarkers endorsed by regulators to enrich clinical trial populations for those at risk of developing T1D. As such, the development of therapies that delay or prevent the onset of T1D remains challenging. To address this drug development need, the Critical Path Institute's T1D Consortium (T1DC) acquired patient-level data from multiple observational studies and used a model-based approach to evaluate the utility of islet AAs as enrichment biomarkers in clinical trials. An accelerated failure time model was developed, discussed in our previous publication, which provided the underlying evidence required to receive a qualification opinion for islet AAs as enrichment biomarkers from the European Medicines Agency (EMA) in March 2022. To further democratize the use of the model for scientists and clinicians, we developed a Clinical Trial Enrichment Graphical User Interface. The interactive tool allows users to specify trial participant characteristics, including the percentage of participants with a specific AA combination. Users can specify ranges for participant baseline age, sex, blood glucose measurement from the 120-minute timepoints of an oral glucose tolerance test, and HbA1c. The tool then applies the model to predict the mean probability of a T1D diagnosis for that trial population and renders the results to the user. To ensure adequate data privacy and to make the tool open-source, a deep learning-based generative model was used to generate a cohort of synthetic subjects that underpins the tool.

Disease progression joint model predicts time to type 1 diabetes onset: Optimizing future type 1 diabetes prevention studies.

Clinical trials seeking type 1 diabetes prevention are challenging in terms of identifying patient populations likely to progress to type 1 diabetes within limited (i.e., short-term) trial durations. Hence, we sought to improve such efforts by developing a quantitative disease progression model for type 1 diabetes. Individual-level data obtained from the TrialNet Pathway to Prevention and The Environmental Determinants of Diabetes in the Young natural history studies were used to develop a joint model that links the longitudinal glycemic measure to the timing of type 1 diabetes diagnosis. Baseline covariates were assessed using a stepwise covariate modeling approach. Our study focused on individuals at risk of developing type 1 diabetes with the presence of two or more diabetes-related autoantibodies (AAbs). The developed model successfully quantified how patient features measured at baseline, including HbA1c and the presence of different AAbs, alter the timing of type 1 diabetes diagnosis with reasonable accuracy and precision (<30% RSE). In addition, selected covariates were statistically significant (p < 0.0001 Wald test). The Weibull model best captured the timing to type 1 diabetes diagnosis. The 2-h oral glucose tolerance values assessed at each visit were included as a time-varying biomarker, which was best quantified using the sigmoid maximum effect function. This model provides a framework to quantitatively predict and simulate the time to type 1 diabetes diagnosis in individuals at risk of developing the disease and thus, aligns with the needs of pharmaceutical companies and scientists seeking to advance therapies aimed at interdicting the disease process.

How can natural language processing help model informed drug development?: a review.

Objective: To summarize applications of natural language processing (NLP) in model informed drug development (MIDD) and identify potential areas of improvement. Materials and Methods: Publications found on PubMed and Google Scholar, websites and GitHub repositories for NLP libraries and models. Publications describing applications of NLP in MIDD were reviewed. The applications were stratified into 3 stages: drug discovery, clinical trials, and pharmacovigilance. Key NLP functionalities used for these applications were assessed. Programming libraries and open-source resources for the implementation of NLP functionalities in MIDD were identified. Results: NLP has been utilized to aid various processes in drug development lifecycle such as gene-disease mapping, biomarker discovery, patient-trial matching, adverse drug events detection, etc. These applications commonly use NLP functionalities of named entity recognition, word embeddings, entity resolution, assertion status detection, relation extraction, and topic modeling. The current state-of-the-art for implementing these functionalities in MIDD applications are transformer models that utilize transfer learning for enhanced performance. Various libraries in python, R, and Java like huggingface, sparkNLP, and KoRpus as well as open-source platforms such as DisGeNet, DeepEnroll, and Transmol have enabled convenient implementation of NLP models to MIDD applications. Discussion: Challenges such as reproducibility, explainability, fairness, limited data, limited language-support, and security need to be overcome to ensure wider adoption of NLP in MIDD landscape. There are opportunities to improve the performance of existing models and expand the use of NLP in newer areas of MIDD. Conclusions: This review provides an overview of the potential and pitfalls of current NLP approaches in MIDD.

Leveraging Real-World Data for EMA Qualification of a Model-Based Biomarker Tool to Optimize Type-1 Diabetes Prevention Studies.

The development of therapies to prevent or delay the onset of type 1 diabetes (T1D) remains challenging, and there is a lack of qualified biomarkers to identify individuals at risk of developing T1D or to quantify the time-varying risk of conversion to a diagnosis of T1D. To address this drug development need, the T1D Consortium (i) acquired, remapped, integrated, and curated existing patient-level data from relevant observational studies, and (ii) used a model-based approach to evaluate the utility of islet autoantibodies (AAs) against insulin/proinsulin autoantibody, GAD65, IA-2, and ZnT8 as biomarkers to enrich subjects for T1D prevention. The aggregated dataset was used to construct an accelerated failure time model for predicting T1D diagnosis. The model quantifies presence of islet AA permutations as statistically significant predictors of the time-varying probability of conversion to a diagnosis of T1D. Additional sources of variability that greatly improved the accuracy of quantifying the time-varying probability of conversion to a T1D diagnosis included baseline age, sex, blood glucose measurements from the 120-minute timepoints of oral glucose tolerance tests, and hemoglobin A1c. The developed models represented the underlying evidence to qualify islet AAs as enrichment biomarkers through the qualification of novel methodologies for drug development pathway at the European Medicines Agency (EMA). Additionally, the models are intended as the foundation of a fully functioning end-user tool that will allow sponsors to optimize enrichment criteria for clinical trials in T1D prevention studies.

Application of machine learning to predict reduction in total PANSS score and enrich enrollment in schizophrenia clinical trials.

Clinical trial efficiency, defined as facilitating patient enrollment, and reducing the time to reach safety and efficacy decision points, is a critical driving factor for making improvements in therapeutic development. The present work evaluated a machine learning (ML) approach to improve phase II or proof-of-concept trials designed to address unmet medical needs in treating schizophrenia. Diagnostic data from the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) trial were used to develop a binary classification ML model predicting individual patient response as either "improvement," defined as greater than 20% reduction in total Positive and Negative Syndrome Scale (PANSS) score, or "no improvement," defined as an inadequate treatment response (<20% reduction in total PANSS). A random forest algorithm performed best relative to other tree-based approaches in model ability to classify patients after 6 months of treatment. Although model ability to identify true positives, a measure of model sensitivity, was poor (<0.2), its specificity, true negative rate, was high (0.948). A second model, adapted from the first, was subsequently applied as a proof-of-concept for the ML approach to supplement trial enrollment by identifying patients not expected to improve based on their baseline diagnostic scores. In three virtual trials applying this screening approach, the percentage of patients predicted to improve ranged from 46% to 48%, consistently approximately double the CATIE response rate of 22%. These results show the promising application of ML to improve clinical trial efficiency and, as such, ML models merit further consideration and development.

Machine Learning in Drug Discovery and Development Part 1: A Primer.

Artificial intelligence, in particular machine learning (ML), has emerged as a key promising pillar to overcome the high failure rate in drug development. Here, we present a primer on the ML algorithms most commonly used in drug discovery and development. We also list possible data sources, describe good practices for ML model development and validation, and share a reproducible example. A companion article will summarize applications of ML in drug discovery, drug development, and postapproval phase.

Pharmacokinetics of Basiliximab for the Prevention of Graft-versus-Host Disease in Patients Undergoing Hematopoietic Cell Transplantation with Minimal-Intensity Cyclophosphamide and Fludarabine.

STUDY OBJECTIVE: Basiliximab is an immunosuppressive monoclonal antibody used for rejection prevention following solid organ transplantation; the pharmacokinetics (PK) of basiliximab in this setting are known. Basiliximab may also be used for prophylaxis and treatment of graft-versus-host disease (GVHD) in patients undergoing allogeneic hematopoietic cell transplantation (HCT); however, the PK of basiliximab in this setting are not known. Clinical transplant providers expect variation in the volume of distribution and clearance after nonmyeloablative allogeneic transplantation (NMAT) compared with solid organ transplantation. Blood loss, organ site-specific antibody accumulation, and differences in blood product use during the two transplantation approaches may generate differences in basiliximab PK. Therefore, the objective of this study was to describe the PK of basiliximab after its addition to a minimally intense NMAT regimen, in conjunction with cyclosporine, for GVHD prophylaxis in patients with hematologic malignancies. DESIGN: Population PK analysis of a single-center, single-arm, phase II clinical trial. SETTING: Academic cancer research center. PATIENTS: Fourteen adults with hematologic malignancies (acute myeloid leukemia, acute lymphoblastic leukemia, chronic lymphocytic leukemia, myelodysplastic syndrome, non-Hodgkin's lymphoma, Hodgkin's lymphoma, myelofibrosis, or severe aplastic anemia) and undergoing NMAT with a fully HLA-matched (10 of 10 antigen matched) related or unrelated donor. MEASUREMENTS AND MAIN RESULTS: Basiliximab was used in conjunction with cyclosporine to deplete activated T cells in vivo as GVHD prophylaxis. We developed a novel competitive enzyme-linked immunosorbent assay (ELISA) method using recombinant interleukin-2 receptor alpha-chain (IL-2Ra) and a commercially available soluble sIL-2R ELISA kit to permit the quantification of serum basiliximab concentrations and characterization of the PK properties of the drug in this patient population. Using a nonlinear mixed effects model with NONMEM software, a one-compartment model with first-order elimination best described the PK, as covariate analysis using stepwise covariate modeling did not improve the base model. CONCLUSION: We suggest a one-compartment population model with first-order elimination to capture the PK profile for basiliximab for this patient population.

The feasibility of genome-scale biological network inference using Graphics Processing Units.

Systems research spanning fields from biology to finance involves the identification of models to represent the underpinnings of complex systems. Formal approaches for data-driven identification of network interactions include statistical inference-based approaches and methods to identify dynamical systems models that are capable of fitting multivariate data. Availability of large data sets and so-called 'big data' applications in biology present great opportunities as well as major challenges for systems identification/reverse engineering applications. For example, both inverse identification and forward simulations of genome-scale gene regulatory network models pose compute-intensive problems. This issue is addressed here by combining the processing power of Graphics Processing Units (GPUs) and a parallel reverse engineering algorithm for inference of regulatory networks. It is shown that, given an appropriate data set, information on genome-scale networks (systems of 1000 or more state variables) can be inferred using a reverse-engineering algorithm in a matter of days on a small-scale modern GPU cluster.

Multiple approaches to predicting oxygen and glucose consumptions by HepG2 cells on porous scaffolds in an axial-flow bioreactor.

In this study, the distribution of oxygen and glucose was evaluated along with consumption by hepatocytes using three different approaches. The methods include (i) Computational Fluid Dynamics (CFD) simulation, (ii) residence time distribution (RTD) analysis using a step-input coupled with segregation model or dispersion model, and (iii) experimentally determined consumption by HepG2 cells in an open-loop. Chitosan-gelatin (CG) scaffolds prepared by freeze-drying and polycaprolactone (PCL) scaffolds prepared by salt leaching technique were utilized for RTD analyses. The scaffold characteristics were used in CFD simulations i.e. Brinkman's equation for flow through porous medium, structural mechanics for fluid induced scaffold deformation, and advection-diffusion equation coupled with Michaelis-Menten rate equations for nutrient consumption. With the assumption that each hepatocyte behaves like a micro-batch reactor within the scaffold, segregation model was combined with RTD to determine exit concentration. A flow rate of 1 mL/min was used in the bioreactor seeded with 0.6 × 10(6) HepG2 cells/cm(3) on CG scaffolds and oxygen consumption was measured using two flow-through electrodes located at the inlet and outlet. Glucose in the spent growth medium was also analyzed. RTD results showed distribution of nutrients to depend on the surface characteristics of scaffolds. Comparisons of outlet oxygen concentrations between the simulation results, and experimental results showed good agreement with the dispersion model. Outlet oxygen concentrations from segregation model predictions were lower. Doubling the cell density showed a need for increasing the flow rate in CFD simulations. This integrated approach provide a useful strategy in designing bioreactors and monitoring tissue regeneration.

Modeling pressure drop using generalized scaffold characteristics in an axial-flow bioreactor for soft tissue regeneration.

The goal of this study was to better understand how analytical permeability models based on scaffold architecture can facilitate a non-invasive technique to real time monitoring of pressure drop in bioreactors. In particular, we evaluated the permeability equations for electrospun and freeze dried scaffolds via pressure drop comparison in an axial-flow bioreactor using computational fluid dynamic (CFD) and experimentation. The polycaprolactone-cellulose acetate fibers obtained by co-axial electrospinning technique and Chitosan-Gelatin scaffolds prepared using freeze-drying techniques were utilized. Initially, the structural properties (fiber size, pore size and porosity) and mechanical properties (elastic modulus and Poisson's ratio) of scaffolds in phosphate buffered saline at 37 °C were evaluated. The CFD simulations were performed by coupling fluid flow, described by Brinkman equation, with structural mechanics using a moving mesh. The experimentally obtained pressure drop values for both 1 mm thick and 2 mm thick scaffolds agreed with simulation results. To evaluate the effect of permeability and elastic modulus on pressure drop, CFD predictions were extended to a broad range of permeabilities spanning synthetic scaffolds and tissues, elastic moduli, and Poisson's ratio. Results indicated an increase in pressure drop with increase in permeability. Scaffolds with higher elastic modulus performed better and the effect of Poisson's ratio was insignificant. Flow induced deformation was negligible in axial-flow bioreactor. In summary, scaffold permeabilities can be calculated using scaffold microarchitecture and can be used in non-invasive monitoring of tissue regeneration.

Modeling of porous scaffold deformation induced by medium perfusion.

In this study, we tested the possibility of calculating permeability of porous scaffolds utilized in soft tissue engineering using pore size and shape. We validated the results using experimental measured pressure drop and simulations with the inclusion of structural deformation. We prepared Polycaprolactone (PCL) and Chitosan-Gelatin (CG) scaffolds by salt leaching and freeze drying technique, respectively. Micrographs were assessed for pore characteristics and mechanical properties. Porosity for both scaffolds was nearly same but the permeability varied 10-fold. Elastic moduli were 600 and 9 kPa for PCL and CG scaffolds, respectively, while Poisson's ratio was 0.3 for PCL scaffolds and ∼1.0 for CG scaffolds. A flow-through bioreactor accommodating a 10 cm diameter and 0.2 cm thick scaffold was used to determine the pressure-drop at various flow rates. Additionally, computational fluid dynamic (CFD) simulations were performed by coupling fluid flow, described by Brinkman equation, with structural mechanics using a dynamic mesh. The experimentally obtained pressure drop matched the simulation results of PCL scaffolds. Simulations were extended to a broad range of permeabilities (10(-10) m(2) to 10(-14) m(2) ), elastic moduli (10-100,000 kPa) and Poisson's ratio (0.1-0.49). The results showed significant deviation in pressure drop due to scaffold deformation compared to rigid scaffold at permeabilities near healthy tissues. Also, considering the scaffold as a nonrigid structure altered the shear stress profile. In summary, scaffold permeability can be calculated using scaffold pore characteristics and deformation could be predicted using CFD simulation. These relationships could potentially be used in monitoring tissue regeneration noninvasively via pressure drop.

Application of computational fluid dynamics in tissue engineering.

The process of tissue regeneration consists of a set of complex phenomena such as hydrodynamics, nutrient transfer, cell growth, and matrix deposition. Traditional cell culture and bioreactor design procedure follow trial-and-error analyses to understand the effects of varying physical, chemical, and mechanical parameters that govern the process of tissue regeneration. This trend has been changing as computational fluid dynamics (CFD) analysis can now be used to understand the effects of flow, cell proliferation, and consumption kinetics on the dynamics involved with in vitro tissue regeneration. Furthermore, CFD analyses enable understanding the influence of nutrient transport on cell growth and the effect of cell proliferation as the tissue regenerates. This is especially advantageous in improving and optimizing the design of bioreactors and tissue culture. Influence of parameters such as velocity, oxygen tension, stress, and strain on tissue growth can be effectively studied throughout the bioreactor using CFD as it becomes impractical and cumbersome to install probes at several locations in the bioreactor. Hence, CFD offers several advantages for the advancement of tissue engineering.

Dynamics of diffusivity and pressure drop in flow-through and parallel-flow bioreactors during tissue regeneration.

In this study, transport characteristics in flow-through and parallel-flow bioreactors used in tissue engineering were simulated using computational fluid dynamics. To study nutrient distribution and consumption by smooth muscle cells colonizing the 100 mm diameter and 2-mm thick scaffold, effective diffusivity of glucose was experimentally determined using a two-chambered setup. Three different concentrations of chitosan-gelatin scaffolds were prepared by freezing at -80°C followed by lyophilization. Experiments were performed in both bioreactors to measure pressure drop at different flow rates. At low flow rates, experimental results were in agreement with the simulation results for both bioreactors. However, increase in flow rate beyond 5 mL/min in flow-through bioreactor showed channeling at the circumference resulting in lower pressure drop relative to simulation results. The Peclet number inside the scaffold indicated nutrient distribution within the flow-through bioreactor to be convection-dependent, whereas the parallel-flow bioreactor was diffusion-dependent. Three alternative design modifications to the parallel-flow were made by (i) introducing an additional inlet and an outlet, (ii) changing channel position, and (iii) changing the hold-up volume. Simulation studies were performed to assess the effect of scaffold thickness, cell densities, and permeability. These new designs improved nutrient distribution for 2 mm scaffolds; however, parallel-flow configuration was found to be unsuitable for scaffolds more than 4-mm thick, especially at low porosities as tissues regenerate. Furthermore, operable flow rate in flow-through bioreactors is constrained by the mechanical strength of the scaffold. In summary, this study showed limitations and differences between flow-through and parallel-flow bioreactors used in tissue engineering.