AI纳米材料创新先驱，释放潜力

Initiation of Coverage AsiaChina Health CarePharmaceuticals /BiotechnologyA pioneer in AI nanomaterial innovation, unleashing potential. BUY Valuation & Risks Cyrus Ng, CFAResearch Analyst+852-2203-6208 We initiate coverage on METiS with a BUY rating and a PT of HK$19.0.As a pioneerin AI-powered nanomaterial technology for drug discovery, METiS is positioned forleadership in AI-driven drug delivery, underpinned by its intellectual property,including an extensive de novo lipid library, advanced AI models, and screeningplatforms. Our PT, derived from a 10-year DCF model (WACC: 10%, terminal growth3%), implies a 26E/27E P/S of 27x/25x. METiS's core platform, NanoForge, integrates AI with quantum chemistry,molecular dynamics, and wet-lab validation.It features the world’s first de novolipid language model, the largest lipid library (over 10 million AI-generatedstructures), and the first end-to-end lipid/LNP screening platform. This closed-loopR&D workflow, powered by the METiS AI Agent, accelerates novel and effectivecandidateidentification through high-confidence virtual screening andexperimental design. Its pipeline is diverse and low-risk,demonstrating innovation and scalable AI-driven platforms. The company's "delivery-first" strategy targets clinically validatedareas, using proprietary nano-delivery to overcome traditional therapy limits. Thisapproach allows pursuit of high-value indications with lower scientific and clinicalrisk. With more than 10 products developed in six years, including one pre-NDAasset, the firm is set for continuous clinical and commercial growth Out-licensing deals and partnerships validate METiS's platform and pipelinevalue.The company recently out-licensed global rights for its preclinical trispecificT-cell engager, MTS-128, for a USD20m upfront payment, with potential milestonesup to USD1.6bn plus royalties. METiS also pursues business development with over30 global companies across diverse therapeutic areas, generating revenue thatfuels platform development and maintains its leadership in drug delivery. Risksrelate to (1) technological dependence and innovation, (2) AI flaws andmisuse, (3) clinical development risks, (4) limited commercialization experience, (5)market competition, and (6) partnership and collaboration. NanoForge — foundationof METiS's AI-drivennanomaterial platforms NanoForgeis Metis's proprietary AI platform,accelerating nanomaterialdiscovery and optimization for targeted drug delivery across therapeutics.It usesproprietaryAI models for molecule generation,prediction,and optimizedformulation. Its architecture integrates a foundational "dry-wet" data layer for end-to-endoptimization, leverages what Frost & Sullivan describe as the world's largest AI-generated ionizable lipid library (over 10 million entries), and utilizes a model layerof proprietary generative and predictive AI (e.g., LipidFLAG, PhatGPT, LipidBERT).All capabilities are accessible via the intelligent METiS Agent, providing traceablerecommendations. Foundational dry-wet data layer — integrated dry-wet lab infrastructure NanoForge's foundational dry-wet data layer accelerates development andimprovesclinical success,driving closed-loop learning and end-to-endoptimization from in-silico design to in-vivo validation.This integrated layerfocuses on two key areas: (1) For lipid optimization, Metis uses mechanistic simulations, AI models, and in- 6 July 2026Pharmaceuticals / BiotechnologyMetis vivo testing to select precise delivery candidates. (2) For formulation optimization, the proprietary METiS Datalots system, an AI-enabledhigh-throughput screening platform,rapidly tests hundreds ofpermutations. This robust infrastructure ensures a continuous feedback loop,quickly refining AI predictions with real-world biological data. Library layer — METiS's ionizable lipid library — METiS's data universe The NanoForge platform features the world's largest ionizable lipid library, withover 10 million AI-generated entries, serving as a core component for targeteddrug delivery.Starting from public structures, Metis employed a fragment-basedapproach and generative AI (including PhatGPT) to create novel lipid structures,addressing diversity gaps. This AI-driven generation is coupled with a continuousfeedback loop from wet-lab validation, where results from over 2,000 animal-testedlipids refine predictive models like LipidBERT and LipidFLAG. Beyond screening,this library acts as a dynamic learning corpus, providing extensive data fordeveloping lipid-specific AI foundation models, much like a large language model,thereby powering the discovery of patentable lipid candidates. Model layer — METiS AI foundational models — generative andpredictive toolkits TheNanoForge platform's model layer comprises a suite of proprietarygenerative and predictive AI foundational models for lipid discovery, LNP, andmRNA design. This includes(1) LipidFLAG, a generative AI model that designs novel, synthesizablelipid