Fast screening of covariates in population models empowered by machine learning

Simulating realistic patient profiles from pharmacokinetic models by a machine learning postprocessing correction of residual variability

We address the problem of model misspecification in population pharmacokinetics (PopPK), by modeling residual unexplained variability (RUV) by machine learning (ML) methods in a postprocessing step after conventional model building. The practical purpose of the method is the generation of realistic virtual patient profiles and the quantification of the extent of model misspecification, by introducing an appropriate metric, to be used as an additional diagnostic of model quality. The proposed methodology consists of the following steps: After developing a PopPK model, the individual residual errors IRES = DV-IPRED, are computed, where DV are the observations and IPRED the individual predictions and are modeled by ML to obtain IRESML. Correction of the IPREDs can then be carried out as DVML = IPRED + IRESML. The methodology was tested in a PK study of ropinirole, for which a PopPK model was developed while a second deliberately misspecified model was also considered. Various supervised ML algorithms were tested, among which Random Forest gave the best results. The ML model was able to correct individual predictions as inspected in diagnostic plots and most importantly it simulated realistic profiles that resembled the real data and canceled out the artifacts introduced by the elevated RUV, even in the case of the heavily misspecified model. Furthermore, a metric to quantify the extent of model misspecification was introduced based on the R2 between IRES and IRESML, following the rationale that the greater the extent of variability explained by the ML model, the higher the degree of model misspecification present in the original model.

Application of machine learning techniques in population pharmacokinetics/pharmacodynamics modeling

Population pharmacokinetics/pharmacodynamics (pop-PK/PD) consolidates pharmacokinetic and pharmacodynamic data from many subjects to understand inter- and intra-individual variability due to patient backgrounds, including disease state and genetics. The typical workflow in pop-PK/PD analysis involves the determination of the structure model, selection of the error model, analysis based on the base model, covariate modeling, and validation of the final model. Machine learning is gaining considerable attention in the medical and various fields because, in contrast to traditional modeling, which often assumes linear or predefined relationships, machine learning modeling learns directly from data and accommodates complex patterns. Machine learning has demonstrated excellent capabilities for prescreening covariates and developing predictive models. This review introduces various applications of machine learning techniques in pop-PK/PD research.

Warfarin-A natural anticoagulant: A review of research trends for precision medication

BACKGROUND

Warfarin is a widely prescribed anticoagulant in the clinic. It has a more considerable individual variability, and many factors affect its variability. Mathematical models can quantify the quantitative impact of these factors on individual variability.

PURPOSE

The aim is to comprehensively analyze the advanced warfarin dosing algorithm based on pharmacometrics and machine learning models of personalized warfarin dosage.

METHODS

A bibliometric analysis of the literature retrieved from PubMed and Scopus was performed using VOSviewer. The relevant literature that reported the precise dosage of warfarin calculation was retrieved from the database. The multiple linear regression (MLR) algorithm was excluded because a recent systematic review that mainly reviewed this algorithm has been reported. The following terms of quantitative systems pharmacology, mechanistic model, physiologically based pharmacokinetic model, artificial intelligence, machine learning, pharmacokinetic, pharmacodynamic, pharmacokinetics, pharmacodynamics, and warfarin were added as MeSH Terms or appearing in Title/Abstract into query box of PubMed, then humans and English as filter were added to retrieve the literature.

RESULTS

Bibliometric analysis revealed important co-occuring MeShH and index keywords. Further, the United States, China, and the United Kingdom were among the top countries contributing in this domain. Some studies have established personalized warfarin dosage models using pharmacometrics and machine learning-based algorithms. There were 54 related studies, including 14 pharmacometric models, 31 artificial intelligence models, and 9 model evaluations. Each model has its advantages and disadvantages. The pharmacometric model contains biological or pharmacological mechanisms in structure. The process of pharmacometric model development is very time- and labor-intensive. Machine learning is a purely data-driven approach; its parameters are more mathematical and have less biological interpretation. However, it is faster, more efficient, and less time-consuming. Most published models of machine learning algorithms were established based on cross-sectional data sourced from the database.

CONCLUSION

Future research on personalized warfarin medication should focus on combining the advantages of machine learning and pharmacometrics algorithms to establish a more robust warfarin dosage algorithm. Randomized controlled trials should be performed to evaluate the established algorithm of warfarin dosage. Moreover, a more user-friendly and accessible warfarin precision medicine platform should be developed.

Machine-Learning Assisted Screening of Correlated Covariates: Application to Clinical Data of Desipramine

Stepwise covariate modeling (SCM) has a high computational burden and can select the wrong covariates. Machine learning (ML) has been proposed as a screening tool to improve the efficiency of covariate selection, but little is known about how to apply ML on actual clinical data. First, we simulated datasets based on clinical data to compare the performance of various ML and traditional pharmacometrics (PMX) techniques with and without accounting for highly-correlated covariates. This simulation step identified the ML algorithm and the number of top covariates to select when using the actual clinical data. A previously developed desipramine population-pharmacokinetic model was used to simulate virtual subjects. Fifteen covariates were considered with four having an effect included. Based on the F1 score (an accuracy measure), ridge regression was the most accurate ML technique on 200 simulated datasets (F1 score = 0.475 ± 0.231), a performance which almost doubled when highly-correlated covariates were accounted for (F1 score = 0.860 ± 0.158). These performances were better than forwards selection with SCM (F1 score = 0.251 ± 0.274 and 0.499 ± 0.381 without/with correlations respectively). In terms of computational cost, ridge regression (0.42 ± 0.07 seconds/simulated dataset, 1 thread) was ~20,000 times faster than SCM (2.30 ± 2.29 hours, 15 threads). On the clinical dataset, prescreening with the selected ML algorithm reduced SCM runtime by 42.86% (from 1.75 to 1.00 days) and produced the same final model as SCM only. In conclusion, we have demonstrated that accounting for highly-correlated covariates improves ML prescreening accuracy. The choice of ML method and the proportion of important covariates (unknown a priori) can be guided by simulations.

Impact of covariate model building methods on their clinical relevance evaluation in population pharmacokinetic analyses: comparison of the full model, stepwise covariate model (SCM) and SCM+ approaches

Covariate analysis in population pharmacokinetics is key for adjusting doses for patients. The main objective of this work was to compare the adequacy of various modeling approaches on covariate clinical relevance decision-making. The full model, stepwise covariate model (SCM) and SCM+ PsN algorithms were compared in a clinical trial simulation of a 383-patient population pharmacokinetic study mixing rich and sparse designs. A one-compartment model with first-order absorption was used. A base model including a body weight effect on CL/F and V/F and a covariate model including 4 additional covariates-parameters relationships were simulated. As for forest plots, ratios between covariates at a specific value and that of a typical individual were calculated with their 90% confidence interval (CI90) using standard errors. Covariates on CL, V and KA were considered relevant if their CI90 fell completely outside the reference area [0.8-1.2]. All approaches provided unbiased covariate ratio estimates. For covariates with a simulated effect, the 3 approaches correctly identify their clinical relevance. However, significant covariates were missed in up to 15% of cases with SCM/SCM+. For covariate with no simulated effects, the full model mainly identified them as non-relevant or with insufficient information while SCM/SCM+ mainly did not select them. SCM/SCM+ assume that non-selected covariates are non-relevant when it could be due to insufficient information, whereas the full model does not make this assumption and is faster. This study must be extended to other methods and completed by a more complex high-dimensional simulation framework.

Machine Learning: A New Approach for Dose Individualization

The application of machine learning (ML) has shown promising results in precision medicine due to its exceptional performance in dealing with complex multidimensional data. However, using ML for individualized dosing of medicines is still in its early stage, meriting further exploration. A systematic review of study designs and modeling details of using ML for individualized dosing of different drugs was performed. We have summarized the status of the study populations, predictive targets, and data sources for ML modeling, the selection of ML algorithms and features, and the evaluation and validation of their predictive performance. We also used the Prediction model Risk of Bias Assessment Tool (PROBAST) to assess the risk of bias of included studies. Currently, ML can be used for both a priori and a posteriori dose selection and optimization, and it can also assist the implementation of therapeutic drug monitoring. However, studies are mainly focused on drugs with narrow therapeutic windows, predominantly immunosuppressants (N = 23, 35.9%) and anti-infectives (N = 21, 32.8%), and there is currently only very limited attention for special populations, such as children (N = 22, 34.4%). Most studies showed poor methodological quality and a high risk of bias. The lack of external validation and clinical utility evaluation currently limits the further clinical implementation of ML for dose individualization. We therefore have proposed several ways to improve the clinical relevance of the studies and facilitate the translation of ML models into clinical practice.

Learning pharmacometric covariate model structures with symbolic regression networks

Efficiently finding covariate model structures that minimize the need for random effects to describe pharmacological data is challenging. The standard approach focuses on identification of relevant covariates, and present methodology lacks tools for automatic identification of covariate model structures. Although neural networks could potentially be used to approximate covariate-parameter relationships, such approximations are not human-readable and come at the risk of poor generalizability due to high model complexity.In the present study, a novel methodology for the simultaneous selection of covariate model structure and optimization of its parameters is proposed. It is based on symbolic regression, posed as an optimization problem with a smooth loss function. This enables training of the model through back-propagation using efficient gradient computations.Feasibility and effectiveness are demonstrated by application to a clinical pharmacokinetic data set for propofol, containing infusion and blood sample time series from 1031 individuals. The resulting model is compared to a published state-of-the-art model for the same data set. Our methodology finds a covariate model structure and corresponding parameter values with a slightly better fit, while relying on notably fewer covariates than the state-of-the-art model. Unlike contemporary practice, finding the covariate model structure is achieved without an iterative procedure involving manual interactions.

Machine Learning in Modeling Disease Trajectory and Treatment Outcomes: An Emerging Enabler for Model-Informed Precision Medicine

The increasing breadth and depth of resolution in biological and clinical data, including -omics and real-world data, requires advanced analytical techniques like artificial intelligence (AI) and machine learning (ML) to fully appreciate the impact of multi-dimensional population variability in intrinsic and extrinsic factors on disease progression and treatment outcomes. Integration of advanced data analytics in Quantitative Pharmacology is crucial for drug-disease knowledge management, enabling precise, efficient and inclusive drug development and utilization - an application we refer to as model-informed precision medicine. AI/ML enables characterization of the molecular and clinical sources of heterogeneity in disease trajectory, advancing end point qualification and biomarker discovery, and informing patient enrichment for proof-of-concept studies as well as trial designs for efficient evidence generation incorporating digital twins and virtual control arms. Explainable ML methods are valuable in elucidating predictors of efficacy and safety of pharmacological treatments, thereby informing response monitoring and risk mitigation strategies. In oncology, emerging opportunities exist for development of the next generation of disease models via ML-assisted joint longitudinal modeling of high-dimensional biomarker data such as circulating tumor DNA and radiomics profiles as predictors of survival outcomes. Finally, mining real-world data leveraging ML algorithms enables understanding of the impact of exclusion criteria on clinical outcomes, thereby informing rational design of appropriately inclusive clinical trials through data-driven broadening of eligibility criteria. Herein, we provide an overview of the aforementioned contexts of use of ML in drug-disease modeling based on examples across multiple therapeutic areas including neurology, rare diseases, autoimmune diseases, oncology and immuno-oncology.

Predictive Modeling of Drug-Related Adverse Events with Real-World Data: A Case Study of Linezolid Hematologic Outcomes

Electronic health records (EHRs) provide meaningful knowledge of drug-related adverse events (AEs) that are not captured in standard drug development and postmarketing surveillance. Using variables obtained from EHR data in the University of California San Francisco de-identified Clinical Data Warehouse, we aimed to evaluate the potential of machine learning to predict two hematological AEs, thrombocytopenia and anemia, in a cohort of patients treated with linezolid for 3 or more days. Features for model input were extracted at linezolid initiation (index), and outcomes were characterized from index to 14 days post-treatment. Random forest classification (RFC) was used for AE prediction, and reduced feature models were evaluated using cumulative importance (cImp) for feature selection. Grade 3+ thrombocytopenia and anemia occurred in 31% of 2,171 and 56% of 2,170 evaluable patients, respectively. Of the total 53 features, as few as 7 contributed at least 50% cImp, resulting in prediction accuracies of 70% or higher and area under the receiver operating characteristic curves of 0.886 for grade 3+ thrombocytopenia and 0.759 for grade 3+ anemia. Sensitivity analyses in strictly defined patient subgroups revealed similarly high predictive performance in full and reduced feature models. A logistic regression model with the same 50% cImp features showed similar predictive performance as RFC and good concordance with RFC probability predictions after isotonic calibration, adding interpretability. Collectively, this work demonstrates potential for machine learning prediction of AE risk in real-world patients using few variables regularly available in EHRs, which may aid in clinical decision making and/or monitoring.

Artificial Intelligence for Quantitative Modeling in Drug Discovery and Development: An Innovation and Quality Consortium Perspective on Use Cases and Best Practices

Recent breakthroughs in artificial intelligence (AI) and machine learning (ML) have ushered in a new era of possibilities across various scientific domains. One area where these advancements hold significant promise is model-informed drug discovery and development (MID3). To foster a wider adoption and acceptance of these advanced algorithms, the Innovation and Quality (IQ) Consortium initiated the AI/ML working group in 2021 with the aim of promoting their acceptance among the broader scientific community as well as by regulatory agencies. By drawing insights from workshops organized by the working group and attended by key stakeholders across the biopharma industry, academia, and regulatory agencies, this white paper provides a perspective from the IQ Consortium. The range of applications covered in this white paper encompass the following thematic topics: (i) AI/ML-enabled Analytics for Pharmacometrics and Quantitative Systems Pharmacology (QSP) Workflows; (ii) Explainable Artificial Intelligence and its Applications in Disease Progression Modeling; (iii) Natural Language Processing (NLP) in Quantitative Pharmacology Modeling; and (iv) AI/ML Utilization in Drug Discovery. Additionally, the paper offers a set of best practices to ensure an effective and responsible use of AI, including considering the context of use, explainability and generalizability of models, and having human-in-the-loop. We believe that embracing the transformative power of AI in quantitative modeling while adopting a set of good practices can unlock new opportunities for innovation, increase efficiency, and ultimately bring benefits to patients.

Low-dimensional neural ODEs and their application in pharmacokinetics

Machine Learning (ML) is a fast-evolving field, integrated in many of today's scientific disciplines. With the recent development of neural ordinary differential equations (NODEs), ML provides a new tool to model dynamical systems in the field of pharmacology and pharmacometrics, such as pharmacokinetics (PK) or pharmacodynamics. The novel and conceptionally different approach of NODEs compared to classical PK modeling creates challenges but also provides opportunities for its application. In this manuscript, we introduce the functionality of NODEs and develop specific low-dimensional NODE structures based on PK principles. We discuss two challenges of NODEs, overfitting and extrapolation to unseen data, and provide practical solutions to these problems. We illustrate concept and application of our proposed low-dimensional NODE approach with several PK modeling examples, including multi-compartmental, target-mediated drug disposition, and delayed absorption behavior. In all investigated scenarios, the NODEs were able to describe the data well and simulate data for new subjects within the observed dosing range. Finally, we briefly demonstrate how NODEs can be combined with mechanistic models. This research work enhances understanding of how NODEs can be applied in PK analyses and illustrates the potential for NODEs in the field of pharmacology and pharmacometrics.

Go beyond the limits of genetic algorithm in daily covariate selection practice

Covariate identification is an important step in the development of a population pharmacokinetic/pharmacodynamic model. Among the different available approaches, the stepwise covariate model (SCM) is the most used. However, SCM is based on a local search strategy, in which the model-building process iteratively tests the addition or elimination of a single covariate at a time given all the others. This introduces a heuristic to limit the searching space and then the computational complexity, but, at the same time, can lead to a suboptimal solution. The application of genetic algorithms (GAs) for covariate selection has been proposed as a possible solution to overcome these limitations. However, their actual use during model building is limited by the extremely high computational costs and convergence issues, both related to the number of models being tested. In this paper, we proposed a new GA for covariate selection to address these challenges. The GA was first developed on a simulated case study where the heuristics introduced to overcome the limitations affecting currently available GA approaches resulted able to limit the selection of redundant covariates, increase replicability of results and reduce convergence times. Then, we tested the proposed GA on a real-world problem related to remifentanil. It obtained good results both in terms of selected covariates and fitness optimization, outperforming the SCM.

On inductive biases for the robust and interpretable prediction of drug concentrations using deep compartment models

Conventional pharmacokinetic (PK) models contain several useful inductive biases guiding model convergence to more realistic predictions of drug concentrations. Implementing similar biases in standard neural networks can be challenging, but might be fundamental for model robustness and predictive performance. In this study, we build on the deep compartment model (DCM) architecture by introducing constraints that guide the model to explore more physiologically realistic solutions. Using a simulation study, we show that constraints improve robustness in sparse data settings. Additionally, predicted concentration-time curves took on more realistic shapes compared to unconstrained models. Next, we propose the use of multi-branch networks, where each covariate can be connected to specific PK parameters, to reduce the propensity of models to learn spurious effects. Another benefit of this architecture is that covariate effects are isolated, enabling model interpretability through the visualization of learned functions. We show that all models were sensitive to learning false effects when trained in the presence of unimportant covariates, indicating the importance of selecting an appropriate set of covariates to link to the PK parameters. Finally, we compared the predictive performance of the constrained models to previous relevant population PK models on a real-world data set of 69 haemophilia A patients. Here, constrained models obtained higher accuracy compared to the standard DCM, with the multi-branch network outperforming previous PK models. We conclude that physiological-based constraints can improve model robustness. We describe an interpretable architecture which aids model trust, which will be key for the adoption of machine learning-based models in clinical practice.

Tumor growth inhibition modeling in patients with second line biliary tract cancer and first line non-small cell lung cancer based on bintrafusp alfa trials

This analysis aimed to quantify tumor dynamics in patients receiving either bintrafusp alfa (BA) or pembrolizumab, by population pharmacokinetic (PK)-pharmacodynamic modeling, and investigate clinical and molecular covariates describing the variability in tumor dynamics by pharmacometric and machine-learning (ML) approaches. Data originated from two clinical trials in patients with biliary tract cancer (BTC; NCT03833661) receiving BA and non-small cell lung cancer (NSCLC; NCT03631706) receiving BA or pembrolizumab. Individual drug exposure was estimated from previously developed population PK models. Population tumor dynamics models were developed for each drug-indication combination, and covariate evaluations performed using nonlinear mixed-effects modeling (NLME) and ML (elastic net and random forest models) approaches. The three tumor dynamics' model structures all included linear tumor growth components and exponential tumor shrinkage. The final BTC model included the effect of drug exposure (area under the curve) and several covariates (demographics, disease-related, and genetic mutations). Drug exposure was not significant in either of the NSCLC models, which included two, disease-related, covariates in the BA arm, and none in the pembrolizumab arm. The covariates identified by univariable NLME and ML highly overlapped in BTC but showed less agreement in NSCLC analyses. Hyperprogression could be identified by higher tumor growth and lower tumor kill rates and could not be related to BA exposure. Tumor size over time was quantitatively characterized in two tumor types and under two treatments. Factors potentially related to tumor dynamics were assessed using NLME and ML approaches; however, their net impact on tumor size was considered as not clinically relevant.

Integrating machine learning with pharmacokinetic models: Benefits of scientific machine learning in adding neural networks components to existing PK models

Recently, the use of machine-learning (ML) models for pharmacokinetic (PK) modeling has grown significantly. Although most of the current approaches use ML techniques as black boxes, there are only a few that have proposed interpretable architectures which integrate mechanistic knowledge. In this work, we use as the test case a one-compartment PK model using a scientific machine learning (SciML) framework and consider learning an unknown absorption using neural networks, while simultaneously estimating other parameters of drug distribution and elimination. We generate simulated data with different sampling strategies to show that our model can accurately predict concentrations in extrapolation tasks, including new dosing regimens with different sparsity levels, and produce reliable forecasts even for new patients. By using a scenario of fitting PK data with complex absorption, we demonstrate that including known physiological structure into an SciML model allows us to obtain highly accurate predictions while preserving the interpretability of classical compartmental models.

Bridging the Worlds of Pharmacometrics and Machine Learning

Precision medicine requires individualized modeling of disease and drug dynamics, with machine learning-based computational techniques gaining increasing popularity. The complexity of either field, however, makes current pharmacological problems opaque to machine learning practitioners, and state-of-the-art machine learning methods inaccessible to pharmacometricians. To help bridge the two worlds, we provide an introduction to current problems and techniques in pharmacometrics that ranges from pharmacokinetic and pharmacodynamic modeling to pharmacometric simulations, model-informed precision dosing, and systems pharmacology, and review some of the machine learning approaches to address them. We hope this would facilitate collaboration between experts, with complementary strengths of principled pharmacometric modeling and flexibility of machine learning leading to synergistic effects in pharmacological applications.

Modeling tumor size dynamics based on real-world electronic health records and image data in advanced melanoma patients receiving immunotherapy

The development of immune checkpoint inhibitors (ICIs) has revolutionized cancer therapy but only a fraction of patients benefits from this therapy. Model-informed drug development can be used to assess prognostic and predictive clinical factors or biomarkers associated with treatment response. Most pharmacometric models have thus far been developed using data from randomized clinical trials, and further studies are needed to translate their findings into the real-world setting. We developed a tumor growth inhibition model based on real-world clinical and imaging data in a population of 91 advanced melanoma patients receiving ICIs (i.e., ipilimumab, nivolumab, and pembrolizumab). Drug effect was modeled as an ON/OFF treatment effect, with a tumor killing rate constant identical for the three drugs. Significant and clinically relevant covariate effects of albumin, neutrophil to lymphocyte ratio, and Eastern Cooperative Oncology Group (ECOG) performance status were identified on the baseline tumor volume parameter, as well as NRAS mutation on tumor growth rate constant using standard pharmacometric approaches. In a population subgroup (n = 38), we had the opportunity to conduct an exploratory analysis of image-based covariates (i.e., radiomics features), by combining machine learning and conventional pharmacometric covariate selection approaches. Overall, we demonstrated an innovative pipeline for longitudinal analyses of clinical and imaging RWD with a high-dimensional covariate selection method that enabled the identification of factors associated with tumor dynamics. This study also provides a proof of concept for using radiomics features as model covariates.

A Systems-Based Analysis of Mono- and Combination Therapy for Carbapenem-Resistant Klebsiella pneumoniae Bloodstream Infections

Antimicrobial resistance is a global threat. As "proof-of-concept," we employed a system-based approach to identify patient, bacterial, and drug variables contributing to mortality in patients with carbapenem-resistant Klebsiella pneumoniae (CRKp) bloodstream infections exposed to colistin (COL) and ceftazidime-avibactam (CAZ/AVI) as mono- or combination therapies. Patients (n = 49) and CRKp isolates (n = 22) were part of the Consortium on Resistance Against Carbapenems in Klebsiella and other Enterobacteriaceae (CRACKLE-1), a multicenter, observational, prospective study of patients with carbapenem-resistant Enterobacterales (CRE) conducted between 2011 and 2016. Pharmacodynamic activity of mono- and combination drug concentrations was evaluated over 24 h using in vitro static time-kill assays. Bacterial growth and killing dynamics were estimated with a mechanism-based model. Random Forest was used to rank variables important for predicting 30-day mortality. Isolates exposed to COL+CAZ/AVI had enhanced early bacterial killing compared to CAZ/AVI alone and fewer incidences of regrowth compared to COL and CAZ/AVI. The mean coefficient of determination (R2) for the observed versus predicted bacterial counts was 0.86 (range: 0.75 - 0.95). Bacterial subpopulation susceptibilities and drug mechanistic synergy were essential to describe bacterial killing and growth dynamics. The combination of clinical (hypotension), bacterial (IncR plasmid, aadA2, and sul3) and drug (KC50) variables were most predictive of 30-day mortality. This proof-of-concept study combined clinical, bacterial, and drug variables in a unified model to evaluate clinical outcomes.

Machine learning advances the integration of covariates in population pharmacokinetic models: Valproic acid as an example

Background and Aim: Many studies associated with the combination of machine learning (ML) and pharmacometrics have appeared in recent years. ML can be used as an initial step for fast screening of covariates in population pharmacokinetic (popPK) models. The present study aimed to integrate covariates derived from different popPK models using ML.

Methods: Two published popPK models of valproic acid (VPA) in Chinese epileptic patients were used, where the population parameters were influenced by some covariates. Based on the covariates and a one-compartment model that describes the pharmacokinetics of VPA, a dataset was constructed using Monte Carlo simulation, to develop an XGBoost model to estimate the steady-state concentrations (Css) of VPA. We utilized SHapley Additive exPlanation (SHAP) values to interpret the prediction model, and calculated estimates of VPA exposure in four assumed scenarios involving different combinations of CYP2C19 genotypes and co-administered antiepileptic drugs. To develop an easy-to-use model in the clinic, we built a simplified model by using CYP2C19 genotypes and some noninvasive clinical parameters, and omitting several features that were infrequently measured or whose clinically available values were inaccurate, and verified it on our independent external dataset.

Results: After data preprocessing, the finally generated combined dataset was divided into a derivation cohort and a validation cohort (8:2). The XGBoost model was developed in the derivation cohort and yielded excellent performance in the validation cohort with a mean absolute error of 2.4 mg/L, root-mean-squared error of 3.3 mg/L, mean relative error of 0%, and percentages within ±20% of actual values of 98.85%. The SHAP analysis revealed that daily dose, time, CYP2C19*2 and/or *3 variants, albumin, body weight, single dose, and CYP2C19*1*1 genotype were the top seven confounding factors influencing the Css of VPA. Under the simulated dosage regimen of 500 mg/bid, the VPA exposure in patients who had CYP2C19*2 and/or *3 variants and no carbamazepine, phenytoin, or phenobarbital treatment, was approximately 1.74-fold compared to those with CYP2C19*1/*1 genotype and co-administered carbamazepine + phenytoin + phenobarbital. The feasibility of the simplified model was fully illustrated by its performance in our external dataset.

Conclusion: This study highlighted the bridging role of ML in big data and pharmacometrics, by integrating covariates derived from different popPK models.

A Multi-Centre Study to Risk Stratify Colorectal Polyp Surveillance Patients Utilising Volatile Organic Compounds and Faecal Immunochemical Test

Simple Summary

A total of 15% of the annual colonoscopies are undertaken for polyp surveillance, and this causes strain on already stretched endoscopy services in the U.K. Hence, there is an urgent need to triage and prioritise patients with high-risk findings. The performance of the faecal immunochemical test is not adequate to be used within polyp surveillance. In light of this, we investigated the usefulness of urinary volatile organic compounds for the detection of high-risk findings. We confirmed that urinary volatile organic compounds can be used to triage patients with high-risk findings. It can be used alone or in combination with faecal immunochemical tests.

Abstract

(1) Background: The service capacity for colonoscopy remains constrained, and while efforts are being made to recover elective services, polyp surveillance remains a challenge. (2) Methods: This is a multi-centre study recruiting patients already on polyp surveillance. Stool and urine samples were collected for the faecal immunochemical test (FIT) and volatile organic compounds (VOC) analysis, and all participants then underwent surveillance colonoscopy. (3) Results: The sensitivity and specificity of VOC for the detection of a high-risk finding ((≥2 premalignant polyps including ≥1 advanced polyp or ≥5 premalignant polyps) were 0.94 (95% CI, 0.88 to 0.98) and 0.69 (95% CI, 0.64 to 0.75) respectively. For FIT, the sensitivity was (≥10 µg of haemoglobin (Hb) / g faeces) 0.54 (95% CI, 0.43 to 0.65) and the specificity was 0.79 (95% CI, 0.73 to 0.84). The probability reduction for having a high-risk finding following both negative VOC and FIT will be 24% if both tests are applied sequentially. (4) Conclusion: The diagnostic performance of VOC is superior to FIT for the detection of a high-risk finding. The performance further improves when VOC is applied together with FIT sequentially (VOC first and then FIT). VOC alone or the combination of VOC and FIT can be used as a triage tool for patients awaiting colonoscopy within a polyp surveillance population, especially in resource-constrained healthcare systems.

Adoption of Machine Learning in Pharmacometrics: An Overview of Recent Implementations and Their Considerations

Pharmacometrics is a multidisciplinary field utilizing mathematical models of physiology, pharmacology, and disease to describe and quantify the interactions between medication and patient. As these models become more and more advanced, the need for advanced data analysis tools grows. Recently, there has been much interest in the adoption of machine learning (ML) algorithms. These algorithms offer strong function approximation capabilities and might reduce the time spent on model development. However, ML tools are not yet an integral part of the pharmacometrics workflow. The goal of this work is to discuss how ML algorithms have been applied in four stages of the pharmacometrics pipeline: data preparation, hypothesis generation, predictive modelling, and model validation. We will also discuss considerations before the use of ML algorithms with respect to each topic. We conclude by summarizing applications that hold potential for adoption by pharmacometricians.

Application of SHAP values for inferring the optimal functional form of covariates in pharmacokinetic modeling

In population pharmacokinetic (PK) models, interindividual variability is explained by implementation of covariates in the model. The widely used forward stepwise selection method is sensitive to bias, which may lead to an incorrect inclusion of covariates. Alternatives, such as the full fixed effects model, reduce this bias but are dependent on the chosen implementation of each covariate. As the correct functional forms are unknown, this may still lead to an inaccurate selection of covariates. Machine learning (ML) techniques can potentially be used to learn the optimal functional forms for implementing covariates directly from data. A recent study suggested that using ML resulted in an improved selection of influential covariates. However, how do we select the appropriate functional form for including these covariates? In this work, we use SHapley Additive exPlanations (SHAP) to infer the relationship between covariates and PK parameters from ML models. As a case-study, we use data from 119 patients with hemophilia A receiving clotting factor VIII concentrate peri-operatively. We fit both a random forest and a XGBoost model to predict empirical Bayes estimated clearance and central volume from a base nonlinear mixed effects model. Next, we show that SHAP reveals covariate relationships which match previous findings. In addition, we can reveal subtle effects arising from combinations of covariates difficult to obtain using other methods of covariate analysis. We conclude that the proposed method can be used to extend ML-based covariate selection, and holds potential as a complete full model alternative to classical covariate analyses.

Contribution of machine learning to tumor growth inhibition modeling for hepatocellular carcinoma patients under Roblitinib (FGF401) drug treatment

Machine learning (ML) opens new perspectives in identifying predictive factors of efficacy among a large number of patients’ characteristics in oncology studies. The objective of this work was to combine ML with population pharmacokinetic/pharmacodynamic (PK/PD) modeling of tumor growth inhibition to understand the sources of variability between patients and therefore improve model predictions to support drug development decisions. Data from 127 patients with hepatocellular carcinoma enrolled in a phase I/II study evaluating once‐daily oral doses of the fibroblast growth factor receptor FGFR4 kinase inhibitor, Roblitinib (FGF401), were used. Roblitinib  PKs was best described by a two‐compartment model with a delayed zero‐order absorption and linear elimination. Clinical efficacy using the longitudinal sum of the longest lesion diameter data was described with a population PK/PD model of tumor growth inhibition including resistance to treatment. ML, applying elastic net modeling of time to progression data, was associated with cross‐validation, and allowed to derive a composite predictive risk score from a set of 75 patients’ baseline characteristics. The two approaches were combined by testing the inclusion of the continuous risk score as a covariate on PD model parameters. The score was found as a significant covariate on the resistance parameter and resulted in 19% reduction of its variability, and 32% variability reduction on the average dose for stasis. The final PK/PD model was used to simulate effect of patients’ characteristics on tumor growth inhibition profiles. The proposed methodology can be used to support drug development decisions, especially when large interpatient variability is observed.

Pharmacometric modeling and machine learning analyses of prognostic and predictive factors in the JAVELIN Gastric 100 phase III trial of avelumab

Avelumab (anti–PD‐L1) is an approved anticancer treatment for several indications. The JAVELIN Gastric 100 phase III trial did not meet its primary objective of demonstrating superior overall survival (OS) with avelumab maintenance versus continued chemotherapy in patients with advanced gastric cancer/gastroesophageal junction cancer; however, the OS rate was numerically higher with avelumab at timepoints after 12 months. Machine learning (random forests, SIDEScreen, and variable‐importance assessments) was used to build models to identify prognostic/predictive factors associated with long‐term OS and tumor growth dynamics (TGDs). Baseline, re‐baseline, and longitudinal variables were evaluated as covariates in a parametric time‐to‐event model for OS and Gompertzian population model for TGD. The final OS model incorporated a treatment effect on the log‐logistic shape parameter but did not identify a treatment effect on OS or TGD. Variables identified as prognostic for longer OS included older age; higher gamma‐glutamyl transferase (GGT) or albumin; absence of peritoneal carcinomatosis; lower neutrophil‐lymphocyte ratio, lactate dehydrogenase, or C‐reactive protein (CRP); response to induction chemotherapy; and Eastern Cooperative Oncology Group performance status of 0. Among baseline and time‐varying covariates, the largest effects were found for GGT and CRP, respectively. Liver metastasis at re‐baseline predicted higher tumor growth. Tumor size after induction chemotherapy was associated with number of metastatic sites and stable disease (vs. response). Asian region did not impact OS or TGD. Overall, an innovative workflow supporting pharmacometric modeling of OS and TGD was established. Consistent with the primary trial analysis, no treatment effect was identified. However, potential prognostic factors were identified.

Population pharmacokinetic model selection assisted by machine learning

A fit-for-purpose structural and statistical model is the first major requirement in population pharmacometric model development. In this manuscript we discuss how this complex and computationally intensive task could benefit from supervised machine learning algorithms. We compared the classical pharmacometric approach with two machine learning methods, genetic algorithm and neural networks, in different scenarios based on simulated pharmacokinetic data. Genetic algorithm performance was assessed using a fitness function based on log-likelihood, whilst neural networks were trained using mean square error or binary cross-entropy loss. Machine learning provided a selection based only on statistical rules and achieved accurate selection. The minimization process of genetic algorithm was successful at allowing the algorithm to select plausible models. Neural network classification tasks achieved the most accurate results. Neural network regression tasks were less precise than neural network classification and genetic algorithm methods. The computational gain obtained by using machine learning was substantial, especially in the case of neural networks. We demonstrated that machine learning methods can greatly increase the efficiency of pharmacokinetic population model selection in case of large datasets or complex models requiring long run-times. Our results suggest that machine learning approaches can achieve a first fast selection of models which can be followed by more conventional pharmacometric approaches.

Application of Machine Learning in Translational Medicine: Current Status and Future Opportunities

The exponential increase in our ability to harness multi-dimensional biological and clinical data from experimental to real-world settings has transformed pharmaceutical research and development in recent years, with increasing applications of artificial intelligence (AI) and machine learning (ML). Patient-centered iterative forward and reverse translation is at the heart of precision medicine discovery and development across the continuum from target validation to optimization of pharmacotherapy. Integration of advanced analytics into the practice of Translational Medicine is now a fundamental enabler to fully exploit information contained in diverse sources of big data sets such as “omics” data, as illustrated by deep characterizations of the genome, transcriptome, proteome, metabolome, microbiome, and exposome. In this commentary, we provide an overview of ML applications in drug discovery and development, aligned with the three strategic pillars of Translational Medicine (target, patient, dose) and offer perspectives on their potential to transform the science and practice of the discipline. Opportunities for integrating ML approaches into the discipline of Pharmacometrics are discussed and will revolutionize the practice of model-informed drug discovery and development. Finally, we posit that joint efforts of Clinical Pharmacology, Bioinformatics, and Biomarker Technology experts are vital in cross-functional team settings to realize the promise of AI/ML-enabled Translational and Precision Medicine.