[Edaily Reporter KIM JI-WAN ] “Going forward, success or failure in AI-driven new drug development will hinge not on having better AI models, but on who secures better data.”

Noh Hae-sung GENINUS Inc.(389030), Head of Research and Development (Chief Technology Officer (CTO), Executive Vice President, Ph.D. in Immunology), gave this response on the 23rd during an interview with E-Daily’s pharmaceutical and biotech premium content platform, Pharm E-Daily, when asked about the future of the AI-driven new drug development industry. He explained that while the emergence of generative AI and large language models (LLMs) has sparked a strong AI boom in the new drug development sector, the core of competitiveness ultimately lies not in algorithms but in how deeply one understands and secures real patient data.

“New Drug Development Is Ultimately More Challenging Than AI”

Recently, AI-driven new drug development has established itself as a key buzzword in the global biotech industry. As the scope of AI applications expands—from candidate compound screening to predicting the likelihood of clinical success—expectations are growing that it can dramatically reduce development time and costs. However, contrary to these expectations, successful cases of large-scale technology exports or commercialization remain limited.

CTO Noh attributed the slower-than-expected progress in the AI-driven drug discovery industry to the inherent complexity of drug development itself.

“No matter how good a target AI identifies, that alone does not guarantee success,” he said. “Numerous processes remain, including candidate optimization, toxicity assessment, pharmacokinetics (PK), pharmacodynamics (PD), clinical trial design, and patient selection strategies.”

He continued, “The earlier the stage of an asset, the more unvalidated factors there are; therefore, even in global technology export agreements, it is common for the proportion of milestone payments at each development stage to be set higher than the upfront payment,” explaining, “AI is a powerful tool that aids drug discovery, but it does not simplify the drug discovery process itself.”

In particular, he drew a clear line against the view that AI will replace the entire drug development process. CTO Noh said, “AI plays a role in aiding decision-making and narrowing down possibilities, but final validation ultimately takes place through experiments and clinical trials. Biology is a much more complex field than people realize.”

“Tactile Sensitivity Is More Important Than a GPT-Level Brain”

To illustrate the future of AI in drug development, CTO Noh cited the rapidly growing physical AI industry as an example.

“These days, humanoid robots can run, jump, and even perform mid-air flips,” he said. “At first glance, they may seem similar to humans, but tasks like picking up a wet cup or naturally unfolding a plastic bag are still not easy for them.”

The reason lies in the difference in sensory perception. While humans move by constantly receiving tactile and pressure information from their skin, muscles, and joints, robots are not yet able to obtain such information sufficiently.

CTO Noh explained, “Simply put, they’re equipped with a GPT-level brain but lack tactile sensitivity in their fingertips,” adding, “Recently, the field of physical AI has also shifted from a competition over larger models to a competition over how precisely it can perceive the world.”

He emphasized that the same applies to AI-driven drug discovery. “Over the past few years, GPU performance and model size were key competitive factors, but going forward, most AI technologies are likely to reach a certain level of standardization,” said CTO Noh. “Ultimately, it is not the AI itself that makes the difference, but the data it learns from.”

“Data Over Models”… Capturing Real-World Patient Biology Through Spatial Omics

CTO Noh pointed out that many AI-driven drug discovery companies currently rely on publicly available genomic data or bulk omics data.

However, actual patient tissue possesses a complexity that is difficult to explain using only averaged data. This is because various cells—including not only cancer cells but also immune cells and fibroblasts—form the tumor microenvironment (TME) and constantly interact with one another.

“No matter how advanced a model is, it cannot learn what it cannot see,” he said. “To truly reflect real-patient biology, we must understand not only cancer cells but also the tumor microenvironment, intercellular interactions, and spatial structures.”

For this reason, GENINUS Inc. has made spatial omics data its core competitive advantage. Spatial omics refers to technology that can analyze not only gene expression levels but also the specific location of a cell within a tissue and which cells it is adjacent to.

The company’s strategy is to bridge the gap between laboratory findings and real-world clinical practice by combining this data with patients’ clinical information and drug response data for AI analysis.

CTO Noh emphasized, “AI analysis that is not linked to patient data is inevitably limited,” adding, “Learning alongside the actual clinical context of patients can lead to meaningful predictions.”

“Companies That Will Survive Are Those with a Proprietary Sensing Layer”

CTO Noh identified the “Sensing

Layer

” as the most critical competitive advantage in the AI-driven new drug development industry going forward. He predicts that companies capable of directly generating and accumulating data—such as spatialomics, cell-cell interactions, time-dependent biological changes, and live-cell dynamics—which was previously difficult to obtain, will ultimately secure a competitive edge.

He noted, “Most of the data currently collected is akin to a snapshot taken at a specific point in time, but actual diseases are dynamic systems that constantly change over time,” adding, “In the future, data that reflects not only spatial information but also the temporal axis will become increasingly important.”

He continued, “While AI models are likely to become similar over time, unique data assets cannot be easily replicated,” adding, “Companies capable of directly observing spatial interactions and patient biology will emerge as the winners in the era of AI-driven drug discovery.”

GENINUS Inc. is also pursuing a leap forward from being an AI platform company to one that possesses its own new drug assets. Currently, the company is utilizing REMED CO.,LTD., its spatial omics-based AI platform, to discover new targets and target pairs. It is implementing a strategy to connect these discoveries to global technology exports.

CTO Noh stated, “Our ultimate goal is to secure meaningful new drug assets and achieve world-class results in technology exports,” adding, “In the long term, we are also considering establishing our own pipeline.”

Meanwhile, CTO Noh graduated from the Department of Biological Sciences at Pohang University of Science and Technology (POSTECH) and earned a Ph.D. in Immunology from the same university’s graduate school. At Samsung Bioepis, he was responsible for antibody drug discovery projects and the establishment of a discovery platform. After serving as Head of the Efficacy Evaluation Team at AMC Science, Mr. Noh currently serves as CTO of GENINUS Inc. He oversees AI-based target identification, the development of new drug discovery platforms, and spatial omics-based new drug discovery strategies.