• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Monday, November 16, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Health

Review Evaluates Open and Closed Artificial Intelligence Across Pharmaceutical Drug Development

by
November 11, 2026
in Health
Reading Time: 5 mins read
0
Review Evaluates Open and Closed Artificial Intelligence Across Pharmaceutical Drug Development

Review Evaluates Open and Closed Artificial Intelligence Across Pharmaceutical Drug Development

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

A comprehensive narrative review authored by Neal M. Davies and Raimar Loebenberg of the University of Alberta evaluates the operational landscape of artificial intelligence across pharmaceutical science, detailing where freely available open-source architectures match proprietary corporate platforms and where commercial suites retain clear structural advantages. Published in the Journal of Pharmacy and Pharmaceutical Sciences, the analysis revisits an editorial framework introduced three years earlier that compared technological adaptation in drug development to evolutionary natural selection. The authors track the transformation of computational methods across the five standard stages of medicine development, encompassing early discovery, preclinical testing, clinical trials, regulatory review, and post-market safety surveillance. Rather than presenting openness as a rigid binary division, the authors conceptualize software access as a spectrum ranging from fully permissive open-source tools to downloadable open-weight weights, free web servers, and closed commercial ecosystems delivered on subscription schedules.

The central conclusion of the evaluation indicates that capable, freely accessible, or open-source software tools now exist for nearly every routine bench research, educational, and laboratory application encountered by pharmaceutical scientists. The authors note that standard computational operations—such as chemical structure depiction, Simplified Molecular Input Line Entry System format parsing, reference retrieval, calibration curve fitting, non-compartmental pharmacokinetic analysis, and basic physiological modelling—no longer require expensive proprietary licenses. However, the review emphasizes that free web services must not be conflated with private or secure local deployments, because online servers typically require the submission of molecular structures or biological sequences to remote third-party systems. In contrast, comprehensive end-to-end drug discovery campaigns and formally audited workflows destined for binding regulatory filings remain largely dominated by established proprietary suites that provide validated software lineages, contractual vendor indemnification, and structured technical support.

Within the initial discovery and early development phase, computational workflows diverge sharply between community-driven repositories and vendor-managed suites across four core tasks: target identification, de novo generative design, virtual molecular docking, and early property prediction. Protein structure prediction represents a prominent milestone, recognized by the 2024 Nobel Prize in Chemistry awarded to Demis Hassabis, John Jumper, and David Baker. While proprietary engines such as Isomorphic Labs IsoDDE pursue commercial drug design, academic laboratories can deploy open-weight or independent open-source models including OpenFold, Chai-1, and Protenix to predict complex biomolecular structures. Newer MIT-licensed open-weight frameworks such as Boltz-2 attempt to predict both macromolecular structure and small-molecule binding affinity, with its developers reporting performance approaching physics-based free-energy perturbation calculations at roughly one thousandth of the computational expense, though the authors emphasize this developer claim originates in an unreviewed preprint requiring independent replication.

Generative chemistry and molecular screening similarly show an expanding open ecosystem operating alongside commercial platforms. Early proofs of concept demonstrated generative models designing potent DDR1 kinase inhibitors within twenty-one days, while open architectures such as REINVENT and transformer-based systems like MolGPT permit academic institutions to execute de novo design pipelines without licensing expenditures. Open cheminformatics toolkits such as RDKit and DeepChem provide programmatic foundations for descriptor generation and format manipulation, supported by chemical databases including PubChem, ChEMBL, and DrugBank. In molecular docking, classical engines like AutoDock Vina, automated servers such as CB-Dock2, and diffusion-based models like DiffDock reframe spatial pose prediction. However, the authors cite independent benchmarking using the PoseBusters evaluation suite showing that deep learning docking engines did not outperform classical physics-based software like AutoDock Vina or CCDC GOLD in terms of physical plausibility or geometric generalization to unseen complexes.

The review details how computational methods are expanding from virtual screening directly into automated wet-laboratory chemical synthesis and pre-formulation design. In 2026, researchers from OpenAI and Molecule.one described an autonomous frontier language model, designated GPT-5.4, coupled to high-throughput robotic instrumentation to optimize the challenging Chan-Lam coupling reaction for primary sulfonamides. Across two microscale screening campaigns involving 10,080 distinct reactions, the model identified TEMPO as a mild additive, raising the mean liquid chromatography estimated product yield from 16.6 percent to 25.2 percent and increasing the proportion of reactions exceeding thirty percent yield from 15.6 percent to 37.5 percent, with benefits confirmed across eleven of fourteen bench-scale substrate pairs. In formulation and absorption modelling, web tools such as FormulationAI, FormulationDE, and FormulationBCS forecast excipient compatibility and Biopharmaceutics Classification System designations, while integrated artificial intelligence physiologically based pharmacokinetic platforms predict human disposition profiles across hundreds of intravenous and oral datasets.

In preclinical and clinical development, the authors note that artificial intelligence primarily supplements rather than displaces empirical wet-laboratory assays, animal pharmacokinetic experiments, and clinical trial cohorts. While standard commercial simulation suites such as Simcyp, GastroPlus, and Phoenix WinNonlin remain routine industry standards, the open-source Open Systems Pharmacology suite comprising PK-Sim and MoBi provides mechanistic physiologically based pharmacokinetic modelling capable of cross-species extrapolation and first-in-human dose estimation. In a published simvastatin case study cited in the review, PK-Sim demonstrated predictive accuracy comparable to Simcyp, supported by an automated qualification framework demonstrating a geometric mean fold error of 1.4 for CYP3A4-mediated metabolic interactions. In clinical trials, population pharmacokinetic-pharmacodynamic modelling tools such as open-source R packages, nlme, and Stan compete with established closed software like NONMEM and Monolix, while machine-learning digital twins developed on historical patient data forecast control arm trajectories to improve statistical power under prognostic covariate adjustments qualified by the European Medicines Agency.

Clinical milestones during the review period illustrate the practical realities and commercial hurdles facing computationally generated medicines entering human trials. Conversely, Exscientia discontinued DSP-1181 after Phase I investigations, underscoring that computational acceleration during lead discovery does not ensure in vivo biological safety or clinical efficacy.

The governance and regulatory oversight of algorithm-informed drug submissions has accelerated internationally, driven by guidelines from major health authorities. In early 2025, the United States Food and Drug Administration issued draft guidance proposing a seven-step risk-based credibility framework anchored directly to a model’s specific context of use, drawing upon experience gained from evaluating more than five hundred pharmaceutical and biologic applications containing computational components since 2016. In parallel, the European Medicines Agency adopted a comprehensive reflection paper on artificial intelligence lifecycle management, followed by joint guidance from the FDA and EMA outlining ten core principles of good practice emphasizing human-centric design, risk proportionality, and transparent contexts of use. Standards bodies have extended existing medical device software concepts, treating unverified third-party machine learning code as software of unknown provenance under IEC 62304, which requires rigorous inventory control, verification, and active monitoring proportionate to prospective patient harm.

To guide working researchers through procurement and implementation decisions, the review formulates an operational definition of model credibility that rejects superficial training-set correlations in favour of proportionate, risk-calibrated evidence. Credibility is dissected across five verifiable dimensions: predictive performance and robustness on external splits, exact computational reproducibility, input data provenance and adherence to FAIR principles, alignment with regulatory credibility standards, and rigorous human oversight. The authors emphasize that human involvement must be deliberately structured rather than treated as a nominal sign-off, distinguishing between human-in-the-loop validation, human-on-the-loop autonomous process supervision, and human-in-command institutional deployment authority. Without rigorous safeguards, domain training, and independent verification logs, automated systems risk inducing pervasive automation bias, where human reviewers passively defer to algorithmically generated outputs or suffer workforce deskilling over extended operational periods.

The review concludes with disciplined projections reaching from the present decade toward 2050, asserting that the enduring competitive advantage for pharmaceutical researchers lies not in master-level proficiency with any single commercial software package, but in cultivating foundational scientific literacy regarding model validation, software provenance, and algorithmic governance. The authors predict that open-weight architectures will progressively match closed systems in general reasoning tasks while automated self-driving laboratories scale design-make-test-analyze cycles in early medicinal chemistry. However, mechanistic causal interpretation, experimental confirmation, and final ethical accountability must remain firmly anchored to human scientists. Ultimately, the authors argue that the foundational challenge in pharmaceutical science is no longer determining whether artificial intelligence should be incorporated into modern drug development workflows, but ensuring that its ongoing implementation remains transparent, auditable, and subject to uncompromising scientific scrutiny.

Subject of Research: Comparative evaluation of open-source and proprietary artificial intelligence tools across pharmaceutical development stages

Article Title: Survival of the fittest tool: choosing between open and closed artificial intelligence in pharmaceutical science

Article References: Davies, N. M., & Loebenberg, R. (2026). Survival of the fittest tool: choosing between open and closed artificial intelligence in pharmaceutical science. Journal of Pharmacy & Pharmaceutical Sciences, 29, Article 17484. https://doi.org/10.3389/jpps.2026.17484

Image Credits: AI Generated

DOI: 10.3389/jpps.2026.17484

Keywords: artificial intelligence, drug development, open-source software, proprietary software, cheminformatics, ADMET prediction, pharmacokinetics, PBPK modelling, molecular docking, protein structure prediction, AlphaFold, PK-Sim, regulatory science, human oversight, model credibility, biopharmaceutics, RDKit, clinical trials

News Source: Blake Davidson. (November 11, 2026). Review Evaluates Open and Closed Artificial Intelligence Across Pharmaceutical Drug Development. Scienmag.

Tags: ADMET predictionAlphaFoldArtificial IntelligencebiopharmaceuticscheminformaticsClinical TrialsDrug developmenthuman oversightmodel credibilitymolecular dockingopen-source softwarePBPK modellingPharmacokineticsPK-Simproprietary softwareProtein structure predictionRDKitregulatory science
Share12Tweet7Share2ShareShareShare1

Related Posts

Researchers Propose Ecosystem-Level Engineering Strategies to Reshape the Human Gut Microbiome

Researchers Propose Ecosystem-Level Engineering Strategies to Reshape the Human Gut Microbiome

November 16, 2026
Systematic Review Evaluates Group Treatments for Anorexia Nervosa

Systematic Review Evaluates Group Treatments for Anorexia Nervosa

November 16, 2026

Dietary Carbohydrate Molecular Structures Regulate Gut Microbiota and Metabolic Health

November 16, 2026

Corneal Gene Expression and Sex Hormones Drive Divergent Remodelling in Keratoconus

November 16, 2026

POPULAR NEWS

  • Single-Cell Sequencing Unveils Immune Signatures Linked to Methotrexate Response in Arthritis

    Single-Cell Sequencing Unveils Immune Signatures Linked to Methotrexate Response in Arthritis

    29 shares
    Share 12 Tweet 7
  • CRISPR Gene Editing Yields Non-Browning Bananas to Extend Shelf Life

    29 shares
    Share 12 Tweet 7
  • Rare Pannonibacter Bloodstream Infection in Hemodialysis Patient Resolves After Catheter Removal

    29 shares
    Share 12 Tweet 7
  • Machine Learning Distinguishes Fenugreek Seeds but Exposes Deep Learning Context Flaws

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Checklist Documents 2,265 South African Spider Species with Nearly Sixty Percent Endemic

Biogenic Nanoparticle Synthesis Offers Sustainable Alternative to Conventional Chemical Production Methods

Researchers Propose Ecosystem-Level Engineering Strategies to Reshape the Human Gut Microbiome

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 87 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.