In contrast, while joint research agreements between AI-based drug development platforms and pharmaceutical companies are becoming increasingly common in South Korea, cases where AI-discovered candidates have actually entered clinical trials can be counted on one hand. Analysts suggest that the gap lies not in AI models or data processing capabilities, but in the organizational resources and capital required to synthesize and validate the compounds proposed by AI, secure Investigational New Drug (IND) approval, and drive the project through clinical development.
According to the pharmaceutical and biotech industry on the 11th, Insilico Medicine, an AI-based new drug development company, began a Phase 3 clinical trial in China on the 7th of last month for “Lentosertib,” a candidate drug for the treatment of idiopathic pulmonary fibrosis (IPF). The trial will be conducted over 52 weeks, recruiting 320 patients from 47 institutions in China. The primary endpoint is the annual rate of decline in forced vital capacity (FVC), a measure of lung function.
The reason Lentosetip is drawing attention is that it goes beyond simply using AI to predict the likelihood of success for existing drug candidates in clinical trials; instead, AI was directly applied to the processes of identifying new drug targets and designing novel compounds. Insilico Medicine identified TNIK as a new target for fibrotic diseases using its proprietary AI platform, “PandaOmics.” Subsequently, using its generative chemistry platform “Chemistry 42,” the company designed new small-molecule compounds that inhibit TNIK and, following actual synthesis and validation, optimized the drug’s efficacy and pharmacokinetic properties. In essence, AI proposed the disease hypothesis and target, and then AI designed compounds tailored to that target.
InSilicoMedicine’s competitive edge extends beyond this single success story. Since 2021, the company has identified 31 preclinical candidates (PCCs) through its AI platform with the goal of advancing them to clinical trials, and 13 of these have received IND approval.
As a result, the speed at which the company generates candidates also differs from traditional new drug discovery methods. For its in-house projects, it took an average of 12 to 18 months from the start of research to the identification of a PCC. During this process, approximately 60 to 200 compounds were actually synthesized and tested per program. Considering that early-stage drug discovery typically takes about 2 years and 6 months to 4 years, this represents a significant reduction in both the trial-and-error involved and the time required during the candidate identification phase.
Strong Interest in AI in Korea, but…
Limited Cases of Advancement to Clinical TrialsWhile there is growing
interestin joint research between
AIplatforms and related companies
in Korea, the number of cases whereactual drug candidates
have advanced to clinical trialsremains
limited. In particular, when compared to Insilico Medicine, the difference is most pronounced not in the scope of AI utilization, but in the experience of repeatedly advancing candidates from multiple pipelines to clinical trials. Insilico Medicine extensively applied AI throughout the development of Lentoceptip—from disease target identification to the design and optimization of new compounds—and has repeatedly applied the same development approach across multiple pipelines.
In South Korea, the most advanced example is Pharos iBio Co., Ltd.’s acute myeloid leukemia (AML) treatment candidate “Rasmotinib” (PHI-101). Rasmotinib, discovered through “Chemiverse”—an AI-based drug discovery platform developed in-house by Pharos iBio Co., Ltd.—has completed a Phase 1 clinical trial in South Korea and Australia involving patients with relapsed or refractory acute myeloid leukemia (AML). Last year, “PHI-501,” a drug candidate for the treatment of intractable solid tumors, also received IND approval for a Phase 1 clinical trial in South Korea.
Oncocross Co.,Ltd.’s sarcopenia candidate “OC514” has also completed a Phase 1 clinical trial in Australia. OC514 is a combination drug that combines two active ingredients; Oncocross Co.,Ltd. utilized its transcriptome-based AI platform, “Raptor AI,” to identify potential new indications such as sarcopenia and cancer-related cachexia. However, it is important to note that OC514 is not a novel compound designed from scratch but rather a drug rediscovery project in which AI was used to identify a combination of existing ingredients and their indications.
In addition, companies such as SyntekaBio,Inc.(226330)and #Standardim are pursuing AI-based candidate discovery and the development of their own pipelines; however, few have achieved the same level of success as Insilico Medicine in repeatedly advancing multiple candidates to the IND and clinical stages using AI.
A researcher at a major biotech company stated, “Communication between organizations that utilize AI in actual research and traditional research teams is often insufficient, so it is generally difficult to accurately gauge the extent to which AI is being utilized even within the company itself.”
Solutions to the “Triple Challenge” of Labor, Capital, and DataThe industry attributes this gap to a lack of capital and specialized personnel. Critics point out that the issue goes beyond simply securing AI developers; it is also difficult to find “interdisciplinary talent” capable of evaluating AI-generated results from a new drug development perspective and translating them into actual synthesis, experimentation, and candidate selection, and companies lack the financial resources to retain such talent over the long term.
In fact, a recent survey conducted by the AI New Drug Research Institute of the Korea Pharmaceutical and Bio-Pharma Association—which surveyed 55 representatives from 44 domestic pharmaceutical and biotech companies, AI specialized firms, and academic and research institutions—found that 89% of respondents said they are already using AI or are considering its adoption. However, 67.3% of respondents also reported that their new drug development departments do not have a single AI specialist. Reasons cited for the inability to secure specialized personnel included a lack of funding and resources, as well as a lack of understanding of AI technology among existing research staff.
Another limitation cited is that investment incentives for pharmaceutical and biotech companies are still insufficient. Since it takes a long time for the effectiveness of AI-driven new drug development to be proven through actual clinical success rates or sales, companies may feel burdened by the need to build up large-scale personnel or infrastructure.
An official from the pharmaceutical and biotech industry stated, “Collaboration in AI-driven new drug development is increasing in Korea, and interest in its importance has grown significantly,” but added, “From a corporate perspective, there is still a lack of a pressing need to invest large-scale personnel and funds to establish the entire process in-house.”
Some experts point out that to enhance competitiveness in AI-driven drug discovery, a national-level integrated data management system for drug development is needed, along with a standardized format that allows for a comprehensive overview of data catalogs. This is because domestic data is often accessible only within closed environments, and data collected for purposes other than drug development frequently requires classification and processing, leading to poor interoperability.
There are also analyses suggesting that the government must introduce support measures, including efforts to attract foreign talent. The United States has already been supporting the development of bio and medical data suitable for AI training, establishing standards, and fostering interdisciplinary talent through the National Institutes of Health (NIH)’s “Bridge2AI” initiative, while providing early-stage research funding to AI-driven new drug development startups through the Small Business Innovation Research (SBIR) and Small Business Technology Transfer (STTR) programs.
An official from a pharmaceutical company stated, “Although relevant executive orders were rescinded after the Donald Trump administration took office, the U.S. had previously been proactive in attracting global talent through executive orders, such as identifying overseas AI talent and shortening visa processing times.” The official added, “If it is difficult to sufficiently train enough professionals in Korea who understand both AI and new drug development in the short term, we should also consider actively recruiting overseas talent with relevant experience to compensate for the shortage of capabilities.”
At the corporate level, companies must establish a structure that includes securing experimental and validation infrastructure—such as synthetic labs—to actually synthesize compounds designed by AI, test their efficacy and toxicity, and then feed those results back into the model for training. A bioindustry official noted, “If it is difficult to establish the facilities in-house, a long-term joint development model is also possible, in which AI companies, pharmaceutical firms, and contract research organizations (CROs) share various rights and responsibilities—from candidate compound discovery through synthesis, validation, preclinical testing, IND filing, and clinical trials.”
