For the first time, pharmaceutical scientists and medicinal chemists can access a unified, benchmark-validated LQM — not as a batch software installation, not as a siloed cheminformatics tool — but as an on-demand molecular intelligence engine delivered as a web service. By submitting a single SMILES string, users receive a comprehensive, 360-degree developability profile of any small molecule in seconds: a capability that previously required hours of computation across multiple platforms and specialized bioinformatics personnel to integrate — removing the computational barriers that have historically limited rigorous molecular characterization to large, well-resourced pharmaceutical organizations, and making that same analytical power available to biotech innovators, academic drug hunters, and rare disease programs worldwide.

What Is a Large Quantitative Model — and Why Does It Matter?

Unlike Large Language Models trained on patterns in text, a Large Quantitative Model is trained on the quantitative language of science itself. The predictBBB™ LQM was built on thousands of characterized drug candidates and small molecules, with their physicochemical properties (thousands of properties or features for each molecule) encoded as vector representations and molecular fingerprints. This architecture enables simultaneous, high-dimensional property prediction at computational speeds orders of magnitude faster than conventional cheminformatics workflows — while preserving the scientific rigor of chemistry-grounded modeling.

The result is a system that does not approximate molecular behavior from literature patterns. It calculates it, in real time, from the underlying quantitative structure of matter — capable of screening tens of thousands of molecules per day through a single, unified web interface. No installation. No data integration overhead. No specialized infrastructure required.

Benchmark-Validated Performance — and a Continued Commitment to Improvement

Lantern's BBB permeability algorithms were contributed to the Therapeutic Data Commons (TDC) leaderboard — one of the most rigorous open benchmarking platforms in computational drug discovery — where five of Lantern's algorithms rank among the top 12 by accuracy and predictive performance. Since that contribution, Lantern has continued to refine and improve the underlying models, with performance advances beyond what is currently reflected on the public leaderboard. The TDC ranking represents a validated baseline — the current state of the platform exceeds it.

A Unified Molecular Intelligence Panel — Delivered in Seconds

The predictBBB™ LQM expands far beyond its origins as a blood-brain barrier permeability predictor. By submitting a single SMILES string through the web interface, researchers receive an exhaustive molecular profile across four integrated dimensions:

Physicochemical Overview: Instant calculation of lipophilicity (logP), polar surface area (TPSA), molecular weight, and ionization state — the properties that determine whether a molecule can reach its target, survive in circulation, and penetrate the right biological membranes.
Comprehensive Drug-Likeness: An integrated scoring panel combining Lipinski's Rule of Five with four additional developability assessments — flagging bioavailability liabilities, metabolic vulnerabilities, and toxicity risk before a molecule ever reaches the lab.
Structural Architecture: Real-time calculation of 25 molecular descriptors covering electronic distribution and functional group composition — enabling chemists to identify which parts of a molecule will be metabolically degraded and to engineer those weaknesses out at the design stage.
Topological Mapping: Quantification of molecular shape, size, and branching complexity to support Structure-Activity Relationship (SAR) modeling — ensuring a candidate's three-dimensional architecture is optimized to fit its intended protein target.
Beyond the BBB: A Universal Engine for All of Small-Molecule Drug Discovery

The name predictBBB.ai reflects where this platform was born — but not the boundaries of what it can do.

While CNS drug development demands exceptionally tight physicochemical control for blood-brain barrier penetration, the underlying calculations are the fundamental parameters of all small-molecule drug design. For non-CNS programs, the platform's utility is arguably broader: the acceptable physicochemical envelope for peripheral targets is wider, and the ability to predict oral bioavailability, P-glycoprotein efflux, intestinal absorption, and hepatic metabolic clearance in real time is equally critical across oncology, cardiovascular medicine, and rare disease indications.

The platform's transporter models — including P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP) predictions — carry particular strategic value in oncology, where tumor overexpression of efflux pumps is a well-established driver of chemotherapeutic resistance. Identifying substrate liability at the molecular design stage, before physical synthesis, is a material competitive advantage that the predictBBB™ web service now makes accessible without infrastructure barriers.

This positions predictBBB.ai as a universal early-stage decision platform — applicable from CNS programs to kinase inhibitors in oncology to novel chemical entities for rare disease indications — and as the first web-native molecular intelligence service of its class grounded in independently benchmarked, quantitative model performance.

"predictBBB.ai began as a CNS permeability predictor — what it has evolved into is a quantitative intelligence engine that speaks the universal language of medicinal chemistry, now accessible to any drug developer in the world as a web service. Five of our core algorithms rank among the top 12 on the Therapeutic Data Commons leaderboard — that is a measurable scientific standard, not a marketing claim — and our team has continued to advance the platform well beyond that baseline. The earliest decisions in a program are the most consequential ones, and we are inviting partners to integrate this capability and work with us to redefine what rational drug design looks like at scale. Developing and making available ground-breaking computational and AI tools for drug development has the potential to introduce new therapies and cures for patients at a velocity that is needed in medicine."

— Panna Sharma, Chief Executive Officer, Lantern Pharma Inc.

Strategic Integration with withZeta.ai

The predictBBB LQM is fully integrated into Lantern's withZeta.ai® multi-agentic AI co-scientist platform, where it directly addresses a persistent challenge in rare oncology: the absence of large historical datasets makes rational, fail-fast molecular design not a preference but a necessity. Within withZeta.ai, the platform enables Lantern and its partners to architect candidates optimized simultaneously for target potency and pharmacokinetic viability — compressing analytical workflows that previously required days of iterative computation into seconds, and embedding that intelligence directly into the broader co-scientist ecosystem.

Part of a Growing Portfolio of Commercially Leverageable AI Assets

The expanded predictBBB LQM represents the latest addition to Lantern Pharma's growing portfolio of proprietary AI technologies developed for drug discovery — a portfolio that includes the RADR® genomic intelligence platform and the withZeta.ai® multi-agentic co-scientist ecosystem. Consistent with Lantern's strategy of building AI assets with both internal pipeline and external commercial value, the Company intends to leverage and monetize these capabilities through partnerships, licensing arrangements, and direct platform access — creating potential revenue streams that complement its clinical development programs.

Collaboration Opportunities and a Subscription-Based Roadmap

Lantern Pharma is actively inviting pharmaceutical companies and biotech innovators to explore early integration of the predictBBB™ LQM web service into their existing discovery workflows. The platform is particularly well-suited for organizations with active small-molecule programs in oncology, CNS, or rare disease seeking to compress hit-to-lead and lead optimization cycle times, reduce dependence on fragmented computational infrastructure, and implement a rigorous, benchmark-validated developability screen at the earliest stages of candidate selection. API-level access for enterprise integration into partner computational pipelines is in active development. Interested organizations are invited to contact Lantern's business development team to initiate a conversation.

This release marks an early milestone in a broader product roadmap for predictBBB.ai. Additional molecular intelligence features and predictive models are planned for integration in the coming months, with the platform expected to launch as part of a subscription-based service for scientists and drug developers globally. Lantern intends for this service to be accessible to the full spectrum of the drug development community — from individual medicinal chemists and academic researchers to enterprise pharmaceutical teams — establishing predictBBB.ai as an ongoing, commercially scalable AI platform asset.