An industry perspective on current QSP trends in drug development.

2023 marks the 10th anniversary of Natpara's submission to the US FDA, which led to the first recorded regulatory interaction where a decision was supported by Quantitative and Systems Pharmacology (QSP) simulations. It had taken about 5 years for the timid QSP discipline to emerge as an effective Model-Informed Drug Development (MIDD) tool with visible impact in the pharmaceutical industry. Since then, the presence of QSP in the regulatory environment has continued to increase, to the point that the Agency reported 60 QSP submissions in 2020 alone, representing ~ 4% of their annual IND submissions [1]. What sort of industry mindset has enabled QSP to reach this level of success? How does QSP fit within the MIDD paradigm? Does QSP mean the same to Discovery and to Clinical Development projects? How do 'platforms' compare to 'fit-for-purpose' QSP models in an industrial setting? Can QSP and empirical Pharmacokinetic-Pharmacodynamic (PKPD) modelling be complementary? What level of validation is required to inform drug development decisions? This article reflects on all these questions, in particular addressing those audiences with limited line-of-sight into the drug industry decision-making machinery.

Quantitative Systems Pharmacology and Machine Learning: A Match Made in Heaven or Hell?

As pharmaceutical development moves from early-stage in vitro experimentation to later in vivo and subsequent clinical trials, data and knowledge are acquired across multiple time and length scales, from the subcellular to whole patient cohort scale. Realizing the potential of this data for informing decision making in pharmaceutical development requires the individual and combined application of machine learning (ML) and mechanistic multiscale mathematical modeling approaches. Here we outline how these two approaches, both individually and in tandem, can be applied at different stages of the drug discovery and development pipeline to inform decision making compound development. The importance of discerning between knowledge and data are highlighted in informing the initial use of ML or mechanistic quantitative systems pharmacology (QSP) models. We discuss the application of sensitivity and structural identifiability analyses of QSP models in informing future experimental studies to which ML may be applied, as well as how ML approaches can be used to inform mechanistic model development. Relevant literature studies are highlighted and we close by discussing caveats regarding the application of each approach in an age of constant data acquisition. SIGNIFICANCE STATEMENT: We consider when best to apply machine learning (ML) and mechanistic quantitative systems pharmacology (QSP) approaches in the context of the drug discovery and development pipeline. We discuss the importance of prior knowledge and data available for the system of interest and how this informs the individual and combined application of ML and QSP approaches at each stage of the pipeline.

Current practices for QSP model assessment: an IQ consortium survey.

Quantitative Systems Pharmacology (QSP) modeling is increasingly applied in the pharmaceutical industry to influence decision making across a wide range of stages from early discovery to clinical development to post-marketing activities. Development of standards for how these models are constructed, assessed, and communicated is of active interest to the modeling community and regulators but is complicated by the wide variability in the structures and intended uses of the underlying models and the diverse expertise of QSP modelers. With this in mind, the IQ Consortium conducted a survey across the pharmaceutical/biotech industry to understand current practices for QSP modeling. This article presents the survey results and provides insights into current practices and methods used by QSP practitioners based on model type and the intended use at various stages of drug development. The survey also highlights key areas for future development including better integration with statistical methods, standardization of approaches towards virtual populations, and increased use of QSP models for late-stage clinical development and regulatory submissions.

Opportunities for Quantitative Translational Modeling in Oncology.

A 2-day meeting was held by members of the UK Quantitative Systems Pharmacology Network () in November 2018 on the topic of Translational Challenges in Oncology. Participants from a wide range of backgrounds were invited to discuss current and emerging modeling applications in nonclinical and clinical drug development, and to identify areas for improvement. This resulting perspective explores opportunities for impactful quantitative pharmacology approaches. Four key themes arose from the presentations and discussions that were held, leading to the following recommendations: Evaluate the predictivity and reproducibility of animal cancer models through precompetitive collaboration. Apply mechanism of action (MoA) based mechanistic models derived from nonclinical data to clinical trial data. Apply MoA reflective models across trial data sets to more robustly quantify the natural history of disease and response to differing interventions. Quantify more robustly the dose and concentration dependence of adverse events through mathematical modelling techniques and modified trial design.

Best Practices to Maximize the Use and Reuse of Quantitative and Systems Pharmacology Models: Recommendations From the United Kingdom Quantitative and Systems Pharmacology Network.

The lack of standardization in the way that quantitative and systems pharmacology (QSP) models are developed, tested, and documented hinders their reproducibility, reusability, and expansion or reduction to alternative contexts. This in turn undermines the potential impact of QSP in academic, industrial, and regulatory frameworks. This article presents a minimum set of recommendations from the UK Quantitative and Systems Pharmacology Network (UK QSP Network) to guide QSP practitioners seeking to maximize their impact, and stakeholders considering the use of QSP models in their environment.

ASPASIA: A toolkit for evaluating the effects of biological interventions on SBML model behaviour.

A calibrated computational model reflects behaviours that are expected or observed in a complex system, providing a baseline upon which sensitivity analysis techniques can be used to analyse pathways that may impact model responses. However, calibration of a model where a behaviour depends on an intervention introduced after a defined time point is difficult, as model responses may be dependent on the conditions at the time the intervention is applied. We present ASPASIA (Automated Simulation Parameter Alteration and SensItivity Analysis), a cross-platform, open-source Java toolkit that addresses a key deficiency in software tools for understanding the impact an intervention has on system behaviour for models specified in Systems Biology Markup Language (SBML). ASPASIA can generate and modify models using SBML solver output as an initial parameter set, allowing interventions to be applied once a steady state has been reached. Additionally, multiple SBML models can be generated where a subset of parameter values are perturbed using local and global sensitivity analysis techniques, revealing the model's sensitivity to the intervention. To illustrate the capabilities of ASPASIA, we demonstrate how this tool has generated novel hypotheses regarding the mechanisms by which Th17-cell plasticity may be controlled in vivo. By using ASPASIA in conjunction with an SBML model of Th17-cell polarisation, we predict that promotion of the Th1-associated transcription factor T-bet, rather than inhibition of the Th17-associated transcription factor RORγt, is sufficient to drive switching of Th17 cells towards an IFN-γ-producing phenotype. Our approach can be applied to all SBML-encoded models to predict the effect that intervention strategies have on system behaviour. ASPASIA, released under the Artistic License (2.0), can be downloaded from http://www.york.ac.uk/ycil/software.

ADME SARfari: comparative genomics of drug metabolizing systems.

MOTIVATION: ADME SARfari is a freely available web resource that enables comparative analyses of drug-disposition genes. It does so by integrating a number of publicly available data sources, which have subsequently been used to build data mining services, predictive tools and visualizations for drug metabolism researchers. The data include the interactions of small molecules with ADME (absorption, distribution, metabolism and excretion) proteins responsible for the metabolism and transport of molecules; available pharmacokinetic (PK) data; protein sequences of ADME-related molecular targets for pre-clinical model species and human; alignments of the orthologues including information on known SNPs (Single Nucleotide Polymorphism) and information on the tissue distribution of these proteins. In addition, in silico models have been developed, which enable users to predict which ADME relevant protein targets a novel compound is likely to interact with.

Modeling of celiac disease immune response and the therapeutic effect of potential drugs.

BACKGROUND: Celiac disease (CD) is an autoimmune disorder that occurs in genetically predisposed people and is caused by a reaction to the gluten protein found in wheat, which leads to intestinal villous atrophy. Currently there is no drug for treatment of CD. The only known treatment is lifelong gluten-free diet. The main aim of this work is to develop a mathematical model of the immune response in CD patients and to predict the efficacy of a transglutaminase-2 (TG-2) inhibitor as a potential drug for treatment of CD. RESULTS: A thorough analysis of the developed model provided the following results:1. TG-2 inhibitor treatment leads to insignificant decrease in antibody levels, and hence remains higher than in healthy individuals.2. TG-2 inhibitor treatment does not lead to any significant increase in villous area.3. The model predicts that the most effective treatment of CD would be the use of gluten peptide analogs that antagonize the binding of immunogenic gluten peptides to APC. The model predicts that the treatment of CD by such gluten peptide analogs can lead to a decrease in antibody levels to those of normal healthy people, and to a significant increase in villous area. CONCLUSIONS: The developed mathematical model of immune response in CD allows prediction of the efficacy of TG-2 inhibitors and other possible drugs for the treatment of CD: their influence on the intestinal villous area and on the antibody levels. The model also allows to understand what processes in the immune response have the strongest influence on the efficacy of different drugs. This model could be applied in the pharmaceutical R&D arena for the design of drugs against autoimmune small intestine disorders and on the design of their corresponding clinical trials.

Relevance of systems pharmacology in drug discovery.

The pharmaceutical industry is in the process of re-inventing its pipeline in an attempt to overcome its increasing phase II and III attrition rates. Here, we describe how systems pharmacology can be used as a risk assessment tool to alleviate this problem before bringing in larger investments. We propose that this translational research tool could provide a valuable, complementary addition to other emerging innovative approaches for target identification and validation in discovery and, ultimately, for aiding clinical trial optimization.

Successful identification of key chemical structure modifications that lead to improved ADME profiles.

The results of a new method developed to identify well defined structural transformations that are key to improve a certain ADME profile are presented in this work. In particular Naïve Bayesian statistics and SciTegic FCFP_6 molecular fingerprints have been used to extract, from a dataset of 1,169 compounds with known in vitro UGT glucuronidation clearance, those changes in chemical structure that lead to a significant increase in this property. The effectiveness in achieving that goal of the thus found 55,987 transformations has been quantified and compared to classical medicinal chemistry transformations. The conclusion is that on average the new transformations found via in silico methods induce increases of UGT clearance by twofold, whilst the classical transformations are on average unable to alter that endpoint significantly in any direction. When both types of transformations are combined via substructural searches (SSS) the average twofold increase in glucuronidation is maintained. The implications of these findings for the drug design process are also discussed, in particular when compared to other methods previously described in the literature to address the question 'Which compound do I make next?'

Structural pairwise comparisons of HLM stability of phenyl derivatives: Introduction of the Pfizer metabolism index (PMI) and metabolism-lipophilicity efficiency (MLE).

Data mining by pairwise comparison of over 150,000 human liver microsome (HLM) intrinsic clearance values stored within the internal Pfizer database has been performed by an automated tool. Systematic probability tables of specific structural changes on the intrinsic clearance of phenyl derivatives have been generated. From these data two new parameters, the Pfizer Metabolism Index (PMI) and Metabolism-Lipophilicity Efficiency (MLE) are introduced for each fragment. The findings are applied to a Topliss style analysis that focuses on metabolic stability.

Novel TOPP descriptors in 3D-QSAR analysis of apoptosis inducing 4-aryl-4H-chromenes: comparison versus other 2D- and 3D-descriptors.

Novel 3D-descriptors using Triplets Of Pharmacophoric Points (TOPP) were evaluated in QSAR-studies on 80 apoptosis-inducing 4-aryl-4H-chromenes. A predictive QSAR model was obtained using PLS, confirmed by means of internal and external validations. Performance of the TOPP approach was compared with that of other 2D- and 3D-descriptors; statistical analysis indicates that TOPP descriptors perform best. A ranking of TOPP>GRIND>BCI 4096=ECFP>FCFP>GRID-GOLPE>>DRAGON>>>MDL 166 was achieved. Finally, in a 'consensus' analysis predictions obtained using the single methods were compared with an average approach using six out of eight methods. The use of the average is statistically superior to the single methods. Beyond it, the use of several methods can help to easily investigate the presence/absence of outliers according to the 'consensus' of the predicted values: agreement among all the methods indicates a precise prediction, whereas large differences between predicted values (for the same compounds by different methods) would demand caution when using such predictions.

Development of in silico models for human liver microsomal stability.

We developed highly predictive classification models for human liver microsomal (HLM) stability using the apparent intrinsic clearance (CL(int, app)) as the end point. HLM stability has been shown to be an important factor related to the metabolic clearance of a compound. Robust in silico models that predict metabolic clearance are very useful in early drug discovery stages to optimize the compound structure and to select promising leads to avoid costly drug development failures in later stages. Using Random Forest and Bayesian classification methods with MOE, E-state descriptors, ADME Keys, and ECFP_6 fingerprints, various highly predictive models were developed. The best performance of the models shows 80 and 75% prediction accuracy for the test and validation sets, respectively. A detailed analysis of results will be shown, including an assessment of the prediction confidence, the significant descriptors, and the application of these models to drug discovery projects.