Daewoong Pharmaceutical said Monday it has established an “AI-based new drug development system,” which preprocesses and databases 800 million molecular models of major compounds that can be immediately utilized for drug development and discovers new drug candidates based on these materials.

The company plans to expand the use of AI to the entire drug development cycle, including preclinical, clinical, and commercialization.

Daewoong Pharmaceutical added that it is making notable research results in obesity, diabetes, and anticancer drugs by combining the DB and drug development system. For example, to develop a drug for obesity and diabetes, it took just two months to discover and optimize an “active substance” that acts on two target proteins simultaneously using its AI system.

"This is an example of AI solving a problem that researchers have been struggling with for over a year," Daewoong Pharmaceutical said.

The company explained that it also took only six months to discover an active substance with cancer cell inhibitory effects using the AI system and secure a patentable “lead substance” through optimization. This project would have taken at least one to two years using the conventional method.

The results of the company's immersion in building an “AI drug development system” over the past two years to solve the challenges of high cost and low efficiency in drug development are becoming visible, it added.

The name Daewoong gave to the DB of molecular models of 800 million compounds that can be purchased and immediately used for new drug development is “DAVID” (Daewoong Advanced Virtual Database). David is a biblical hero who defeated Goliath in a single blow. The name reflects the researchers' determination to compete with global big pharma in the drug discovery race with AI.

The company noted that the figure of 800 million is a combination of compounds that Daewoong Pharmaceutical has secured through drug research over the past 40 years and almost all compounds currently available for drug development. However, the open-source compounds disclosed globally are not suitable as data for AI drug discovery.

Preprocessing, which separates and removes unnecessary information from complex compound structures, is essential. AI researchers at Daewoong prioritized this task and processed all the data into AI-usable data. Only then was it possible to conduct “AI drug discovery” based on 800 million compounds.

The data of 800 million compounds is the nourishment for AI to grow. Without quality data, AI is useless. The company said that by building the database, Daewoong has laid the groundwork for a quantum jump in the AI drug discovery race.

Today, researchers estimate that the number of compounds that can become new drug candidates is about the power of 10 to 60. It is an unknown number beyond what humans can express in words. The 800 million compounds secured by Daewoong Pharmaceutical are nine powers of 10.

"The world of drug candidates is like the universe, and it is no exaggeration to say that AI has opened a new era of drug discovery," said Park Jun-seok, head of the Center for Drug Discovery. "If we pioneer the unknown with AI, we will be able to develop a large number of new drug candidates and excellent new drugs faster, making a significant contribution to human health."

After building a large database for AI drug discovery, Daewoong developed an AI-based Virtual Screening (AIVS) tool that can be applied to the first stage of drug discovery. This system uses AI to discover 'active substances' against target proteins, can be explored in various ways based on 3D modeling, and can quickly find new patentable active substances with the same chemical properties using generative AI.

Based on these databases and tools, Daewoong opened the AI drug discovery system “DAISY” (Daewoong AI System) last year. The system is a web-based “AI drug discovery portal” where Daewoong researchers can access DAISY to discover new compounds and quickly predict their drug properties.

Even the so-called ADMET research is possible with AI. ADMET stands for Absorption, Distribution, Metabolism, Excretion, and Toxicity and is a research step that identifies the drug properties of a compound, including absorption, distribution, metabolism, excretion, and toxicity. It is a very important step in the early stages of drug development, and if this research is not done properly, the drug is likely to fail in the clinic.

Daewoong Pharmaceutical's researchers have been using AIVS, an AI drug discovery tool, for just a few months and have seen tangible results. In addition to achievements in the fields of obesity, diabetes, and anticancer drugs, they have also achieved meaningful results in the development of proteolytic drugs and have dramatically reduced the trial and error of researchers by simultaneously designing antibodies and evaluating their stability. By discovering and designing candidate substances using AI, they shorten the “time” for new drug development.

"It would be a mistake to look at AI as a technology that replaces humans,” Park said. "Just as deep learning AI learns and grows by accumulating ‘data,’' researchers can only take a step closer to the success of new drug development if they grow together by increasing their ‘insights.’”

According to the Korea Health Industry Promotion Institute, drug development takes an average of 15 years, and only one out of every 10,000 candidates is successful. It takes an average of five years for researchers to find the first drug candidate and another two years to select candidates for clinical trials. From there, it takes another six years to get through phases 1,2 and 3 clinical trials to find one meaningful discovery. It takes another two years to get U.S. FDA approval to go global. That's a long 15-year process, and that's in the best-case scenario.

The “Food and Drug R&D Issues Report” published by the Ministry of Food and Drug Safety last November also assumes a 15-year development period and 2 to 3 trillion won ($1.5 to $2.2 billion) in development costs for a new drug in the United States. The report analyzed if AI technology is applied, the development period can be reduced to seven years, and the cost to about 600 billion won.