[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-121989-en":3,"doc-seo-121989-105":30,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},121989,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",6,"Technology","Machine Learning and Artificial Intelligence in PK-PD Modeling: Fad, Friend, or Foe?","Developing pharmacokinetic-pharmacodynamic (PK-PD) models requires extensive time and specialized expertise, creating a supply-demand gap. The document evaluates how machine learning (ML) and artificial intelligence (AI) could reduce manual supervision effort while improving predictive performance, but it highlights current limitations. It structures key research trends and defines boundaries and opportunities for PK-PD use, emphasizing when AI may substitute conventional models and when hybrid approaches are needed to preserve biological plausibility and reliable out-of-sample predictions.","PERSPECTIVES  \nPERSPECTIVE  \nMachine Learning and Artificial Intelligence in PK-PD Modeling: Fad, Friend, or Foe?  \nZhonghui Huang1, Paolo Denti2  , Hitesh Mistry3  and Frank Kloprogge4,*   \nDeveloping pharmacokinetic-pharmacodynamic (PK-PD) models requires a significant amount of time from highly skilled scientists and the demand for this expertise far outstrips the current supply. The use of machine learning (ML) and artificial intelligence (AL) in PK-PD modeling promises to reduce the number human supervision hours and improve predictive performance, but in its current form it suffers from various limitations. In this perspective, we aimed to structure the main trends and define boundaries and opportunities.  \nMACHINE LEARNING AND PK AND PK-PD MODEL DEVELOPMENT  \nPK-PD modelers leverage various AI-based tools in their daily routine and this has without doubt contributed to increased productivity. For example, AI-driven programming tools can help generating code, for example, Git-Hub Co-Pilot, presenting results, for example, Microsoft Co-Pilot, and scientific writing, such as Large Language models. However, unlike in other fields, AI/ML has not yet been integrated into a PK-PD modeler’s daily routine for analysis of time series data.  \nCurrently, most efforts present proof of concepts, focused on adopting methodologies used in other fields. Broadly speaking, current applications of ML to PK-PD modeling can be divided into two  \nmain streams. The first aims to predict individual observations, exposures, etc., using purely data-driven AI algorithms, that is, using AI as a substitute for the conventional model-based approach encompassing a structural model, variability, and covariate effects. The second does not substitute the conventional model-based approach, but it rather aims at using AI algorithms to guide, inform, and expedite the development of nonlinear mixed-effect model structures and parameter-covariate relations. Thus, the focus is on increasing efficiency and not moving away from the familiar models we currently use (e.g., ordinary differential equations or analytical solutions thereof ), which are grounded in our knowledge of biology, physiology, and pharmacology.  \nHere, we aimed to structure the main trends in published AI/ML research, applied to PK-PD, and define present boundaries and opportunities.  \nML AS SUBSTITUTE OF CONVENTIONAL STRUCTURAL AND STOCHASTIC MODELS  \nAlgorithms that are members of the neural network and tree-based families are reported to display good fits for individual observations in PK and PD time series data. Likewise, it is commonly known that spline models fit PK and PD time series data well.  \nWhat both aforementioned ML and spline models have in common is that they are driven solely by the experimental data they are fit on. There is a complete absence of a mechanistic or semimechanistic structure relatable to conceptual biological knowledge, such as the primary PK concepts of absorption, distribution, and elimination. There is, however, broad consensus that having these conceptual biological priors embedded in our PK-PD models enables meaningful characterization of the variability between patients. Kreutzner et al. 2022, for instance, showed that AI methods performed inferior to conventional population PK (PopPK) models in quantifying random variability for their example.1  \nFurthermore, because no plausibility and consistence with physiology, biology, or pharmacology is embedded in these types of ML models, out of sample predictions are unreliable. For example, prediction of concentrations at times beyond those observed in the clinical study or simulation of exposure at dosing regimens that were not studied appear to be a challenge.2,3 Until this issue is addressed, ML algorithms describing PK and PD time series data cannot be used for dose/schedule optimization or  \n1Great Ormond Street Institute of Child Health, University College London, London, UK; 2Division of Clinical Pharmac","cbCaijaVtEv0uwIe","https://ap.wps.com/l/cbCaijaVtEv0uwIe","pdf",98914,1,3,"English","en",105,"# Machine Learning and Artificial Intelligence in PK-PD Modeling: Fad, Friend, or Foe?\n## ML and PK-PD model development workflows\n## Two main application streams in PK-PD\n## ML as a substitute for conventional structural and stochastic models\n## ML-assisted modeling for structural and stochastic selection","[{\"question\":\"Why are ML and AI considered for PK-PD modeling?\",\"answer\":\"They aim to reduce the time and human supervision needed and potentially improve predictive performance, but current methods face important limitations.\"},{\"question\":\"What two main streams describe current ML applications in PK-PD modeling?\",\"answer\":\"One stream uses purely data-driven AI to predict observations/exposures as a substitute for model-based approaches. The second stream keeps conventional model-based frameworks and uses AI to guide, inform, and expedite developing nonlinear mixed-effect structures and parameter-covariate relations.\"},{\"question\":\"What is a core limitation of ML models that purely fit experimental data?\",\"answer\":\"They often lack mechanistic or semimechanistic biological structure, so physiological plausibility is not embedded; consequently, out-of-sample predictions can be unreliable for unobserved times or dosing regimens not studied.\"}]","Machine Learning and Artificial Intelligence in PK-PD Modeling: Fad, Friend, or Foe? | PDF",1785808172,8,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":28},"machine-learning-and-artificial-intelligence-in-pk-pd-modeling-fad-friend-or-foe","",{"@graph":36,"@context":84},[37,53,67],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":21},"https://docshare.wps.com/document/technology/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/machine-learning-and-artificial-intelligence-in-pk-pd-modeling-fad-friend-or-foe/121989/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-08-04",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"Why are ML and AI considered for PK-PD modeling?","Question",{"text":74,"@type":75},"They aim to reduce the time and human supervision needed and potentially improve predictive performance, but current methods face important limitations.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"What two main streams describe current ML applications in PK-PD modeling?",{"text":79,"@type":75},"One stream uses purely data-driven AI to predict observations/exposures as a substitute for model-based approaches. The second stream keeps conventional model-based frameworks and uses AI to guide, inform, and expedite developing nonlinear mixed-effect structures and parameter-covariate relations.",{"name":81,"@type":72,"acceptedAnswer":82},"What is a core limitation of ML models that purely fit experimental data?",{"text":83,"@type":75},"They often lack mechanistic or semimechanistic biological structure, so physiological plausibility is not embedded; consequently, out-of-sample predictions can be unreliable for unobserved times or dosing regimens not studied.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,112,117,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":110,"slug":111},50,"technology",{"id":113,"doc_module":4,"doc_module_name":46,"category_name":114,"show_sort_weight":115,"slug":116},7,"Healthcare",40,"healthcare",{"id":29,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},"Research & Report",30,"research-report",{"id":122,"doc_module":4,"doc_module_name":46,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":46,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":46,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]