Galaxy and Seoul National University Develop Immunogenicity Prediction Model 'T-SCAPE'

Published in the International Journal "Science Advances"

Galaxy, an AI drug development company, announced on December 11 that it has developed an artificial intelligence (AI) model called 'T-SCAPE' in collaboration with a research team from Seoul National University. This model precisely predicts the T-cell immunogenicity of drug candidates, and the results of this research have been published in the international journal 'Science Advances.'

Galaxy

Immunogenicity is one of the major risk factors considered in the development of protein-based therapeutics. It can lead to reduced efficacy or strong immune responses against the therapeutic agent. However, due to the lack of relevant data and the complexity of immune mechanisms, quantitatively predicting immunogenicity remains a significant challenge.

T-SCAPE is designed to maximize predictive power even when direct data on immunogenicity is limited, by integratively learning from a variety of immunology-related datasets. By connecting different types of biological data-including human and non-human peptide sequences, MHC binding information, T-cell receptor (TCR) interactions, and T-cell activation experiment data-the model is able to capture complex patterns that are difficult for single-dataset models to identify.

Dr. Noh Jinseong of Galaxy, who co-led the study, explained, "Because there is an absolute lack of direct data on immunogenicity, we applied a 'pre-training' strategy in which the AI first learns biological principles." He added, "In particular, we significantly improved predictive performance by applying a deep learning methodology called 'adversarial domain adaptation,' which reduces differences between disparate datasets and identifies common rules."

Validation results showed that T-SCAPE achieved top-tier accuracy in key benchmark assessments for predicting the immunogenicity of peptide-MHC complexes (pMHC).

Seok Chaok, CEO of Galaxy, stated, "This study is significant because, in addition to our AI-based precision protein design capabilities, it establishes a reference point for preemptively evaluating immune responses to therapeutics." He added, "Through this, we will continue to develop our technology to reduce uncertainty and trial-and-error in the process of securing drug candidates and to enhance the efficiency of drug development."