Hanmi Pharmaceutical Presents Research on Proprietary AI Platform 'HARP-pSAR' at ISMB 2026 in the US

Hanmi Pharmaceutical announced on the 24th that it has secured next-generation drug development technology by leveraging its proprietary artificial intelligence (AI) platform to predict the pharmacological activity resulting from minute changes in protein sequences and to suggest optimal directions for candidate substance design.

Particularly noteworthy is the fact that a domestic pharmaceutical and biotech company has discovered a new 'first-in-class' candidate substance using AI and advanced it to the clinical stage for global development, positioning itself as a leading case in the industry.

Seungsu Han, Senior Researcher of Future Growth Division at Hanmi Pharmaceutical, explained to attendees at ISMB 2026 held on the 16th (local time) in Washington D.C., USA, based on a poster containing research on deriving an obesity new drug pipeline based on muscle strengthening using the self-developed AI platform ‘HARP-pSAR’. Hanmi Pharmaceutical

Utilizing its proprietary AI platform, Hanmi Pharmaceutical has derived the world’s first “next-generation obesity innovative drug” (LA-UCN2, code name: HM17321), which enables simultaneous weight loss and muscle gain, and is rapidly advancing its clinical development. Last month, the company also succeeded in securing a next-generation muscle-enhancing therapy (LA-MSTN, code name: HM500197) that is based on a world-first peptide-driven myostatin inhibition mechanism, which was unveiled at the American Diabetes Association (ADA).

Hanmi Pharmaceutical, the core business unit of Hanmi Science, announced on the 24th that it presented research results at the International Conference on Intelligent Systems for Molecular Biology (ISMB), held in Washington D.C., USA from July 12 to 16. The research utilized its in-house developed AI platform “HARP-pSAR” to derive a new drug pipeline for obesity treatment based on muscle enhancement.

ISMB is the world’s largest conference in bioinformatics and computational biology and is recognized as a leading academic society advancing AI-driven research in biology and new drug development. This year, the conference set AI’s competitiveness benchmarks as the ability to solve real-world research problems and its connectivity to drug development outcomes, emphasizing AI’s role not merely as a predictive tool but as a ‘research collaborator.’

The research Hanmi Pharmaceutical presented at this conference attracted significant attention, not only because it demonstrated the prediction performance of its AI model but also because the protein sequence designs proposed by AI were experimentally validated, optimized as candidate substances, and successfully advanced to clinical-stage innovative drugs.

The next-generation obesity treatment, HM17321, is a first-in-class drug that goes beyond simply compensating for muscle loss by enabling the previously considered impossible: simultaneous muscle mass gain and selective fat reduction. The product is currently progressing smoothly in Phase 1 clinical trials in the United States.

HM17321 is designed as a UCN2 (Urocortin-2) analog that selectively targets the CRF2 (corticotropin-releasing factor 2) receptor, rather than GLP-1 or other incretin receptors. According to Hanmi Pharmaceutical, the CRF family acts as signaling molecules related to stress response and recovery, and by selectively targeting the CRF2 receptor, it is possible to directly induce fat reduction, muscle growth, and improved muscle function.

The key challenge in developing HM17321 was to finely optimize the sequence of the UCN2 analog so that it would bind strongly to its target receptor CRFR2 while exhibiting no activity towards the non-target receptor CRFR1.

Building on its unique drug design capabilities, Hanmi Pharmaceutical dramatically enhanced candidate discovery and validation processes by employing its proprietary AI platform, “HARP-pSAR (Hanmi AI-driven Research Platform – protein Sequence-Activity Relationship).”

HARP-pSAR is notable for overcoming the major bottleneck in early-stage drug development—insufficient experimental data (“Low-N”)—and thereby maximizing research efficiency. Unlike traditional AI models that require massive datasets and high-performance computing infrastructure, HARP-pSAR employs quantization techniques to reduce computational costs and is designed for seamless, rapid operation even in local workstation environments.

With only several dozen internally generated experimental results, Hanmi Pharmaceutical built a model capable of accurately predicting activity against both target and non-target receptors. This streamlined what had previously been a labor-intensive process of individually designing, synthesizing, and testing each candidate protein sequence for activity, and enabled ultra-fast screening to identify promising drug candidates based on AI-driven predictions.

In particular, HARP-pSAR is recognized as a demonstrable case of greatly improving the development completeness of HM17321 and dramatically accelerating its clinical entry, as it can predict the activity of novel sequences not included in training data and suggest optimal positions for neo-residues (unknown amino acid residues), thereby significantly reducing unnecessary trial-and-error and experimental costs in the candidate design process.

This approach aligns with the latest trend of “AI-native drug development,” in which results proposed by generative AI or foundation models are experimentally validated and then fed back into the learning process. Hanmi Pharmaceutical plans to further advance HARP-pSAR beyond a single-task prediction model and transform it into a drug design platform technology continuously expandable according to target characteristics.

Choi Inyoung, Executive Vice President and Head of the Future Growth Division at Hanmi Pharmaceutical, stated, “This research is a pioneering case showing that AI can become a practical research partner that goes beyond simple data analysis tools, combining with researchers’ expertise to guide candidate drug design and reduce trial-and-error throughout the development process. In particular, it is highly significant that at ISMB, where world-renowned AI and computational biology researchers gather, Hanmi had the opportunity to present the successful translation of AI-based design outcomes into real drug candidates and clinical development.”

This content was produced with the assistance of AI translation services.