There are cases showing that AI technology can lower time and cost barriers in the early stages of new drug development. SK Telecom and SK Biopharmaceuticals announced that they have discovered candidate substances for targeted treatments for intractable cancer through AI-based joint research and succeeded in reducing the initial research period, which typically takes 1 to 2 years, to about 5 months.

SK Telecom announced on the 15th that it has secured two initial active substances (hits) that showed the potential to bind to the cancer cell surface protein 'ROR1' in an AI-based drug discovery study conducted jointly with SK Biopharmaceuticals.

ROR1 is a tumor-associated cell surface protein that is overexpressed in various types of cancer, such as blood cancer and solid tumors (cancers that form as lumps in organs or tissues), and is a target of interest in the field of anticancer targeted therapy research.

In this study, SK Biopharmaceuticals established a strategy for discovering new binders based on its experience in new drug development.

A binder is a substance designed to bind to specific targets, such as cancer cells, and must satisfy complex conditions including target binding affinity and structural stability.

SKT generated a large number of new binder candidates by applying machine learning that combines and expresses protein fragments in various ways.

Here, reinforcement learning (RL) was combined to induce the search for the optimal structure by assigning high rewards to combinations with high structural stability.

The two companies explained that this approach complements the limitation that research exploring new material structures faces the challenge of securing sufficient data for AI training, which inevitably restricts the range of candidate search using existing data-dependent methods.

In the candidate selection stage, multiple candidates were processed in parallel using SKT's GPU resources, and the combination structure and possibility with ROR1 were rapidly predicted and analyzed through an AI model to narrow down the targets for laboratory verification.

In actual laboratory verification, two of these were as initial active substances It was confirmed that it shows potential.

The two companies stated that the research period took about five months, which is more than 60% shorter than SK Biopharmaceuticals' existing method.

Cho Dong-yeon, Head of AI Convergence at SK Telecom, said, “Based on this achievement, we are also considering expanding the scope of technological cooperation across the entire bio AI field, including the development of bio-specialized LLMs (Large Language Models) utilizing our proprietary AI foundation model.”