Korea Institute of Science and Technology (KIST) said on the 1st that the research team led by principal researcher Park Geun-wan of the natural product drug development division won the international drug candidate prediction competition using artificial intelligence (AI), the CACHE challenge, for the second consecutive time. After tying for first place in last year's competition that searched for drug candidates for the novel coronavirus disease (COVID-19), they achieved sole first place in this competition.
CACHE is an English acronym meaning computer-aided assessment of the suitability of hit-finding experiments for drug discovery. The CACHE challenge, launched in 2021, is an international competition that objectively evaluates how effectively AI technology works in the first stage of drug development.
In particular, it does not remain a simple theory or computer simulation (mock experiment); the drug candidates proposed by participating teams are validated in actual laboratories. Because universities, research institutions and pharmaceutical companies around the world participate and compete, AI drug technologies can gain international recognition.
The fourth competition focused on predicting drug candidates targeting cancer immunotherapy and autoimmune diseases, and 23 teams from around the world advanced to the finals and competed over two years. A total of 1,688 drug candidates were selected for testing, and among them only the candidates proposed by the KIST team were recognized for both chemical originality and biological activity.
Using ECBS (Evolutionary Chemical Binding Similarity), an AI model developed in-house by KIST, the team led by principal researcher Park Geun-wan identified drug candidates targeting the CBLB protein, which has attracted attention as a key target for cancer immunotherapy.
The CBLB protein helps cancer cells evade attacks by the body's immune cells. Targeting this protein can induce a stronger immune response against cancer cells. The ECBS technology is a platform that can quickly predict actual effects by analyzing the structure and characteristics of candidate compounds, and in this competition it was recognized for both predictive precision and practicality.
Principal researcher Park said, We entered the CACHE challenge four times in total to validate the performance of the AI model developed in-house at KIST, and we advanced to the finals consecutively, adding, This victory is the result of externally proving our research achievements. He noted that although various drug development methods have emerged thanks to recent advances in AI technology, the success rate of finding candidate compounds is very low, around 1 to 3 percent, and added, We will continue technology development and strengthen global drug development competitiveness through cooperation with domestic and international pharmaceutical companies.