3billion Named Best FGFR Mutation Prediction Team in CAGI7 Genome Interpretation AI Competition

3billion , an AI-based rare disease diagnosis company, announced on the 15th that it was selected as the best team in the FGFR mutation function prediction challenge at the CAGI7 conference held in Boston, USA from December 6th to 8th and presented its research results.

Hosted by the National Institutes of Health and UC Berkeley, CAGI is an international competition for AI in genome interpretation. It serves as a global platform for blindly evaluating the impact of genetic mutations on disease and protein function. It serves as an international benchmark for academia and industry to compare and validate the performance of AI models under similar conditions.

3billion previously demonstrated its global competitiveness by being selected as the top team in the whole-genome-based rare disease diagnosis and protein function prediction task at CAGI6 in 2022. This year, at CAGI7, the company participated in a challenging task: predicting the impact of all amino acid substitution mutations in genes ranging from FGFR1 to FGFR4 on protein function.

FGFR genes are known to be associated with various diseases, including cancer. This challenge focused on assessing the ability to accurately predict how individual genetic mutations alter protein function, rather than specific diseases. In clinical and research settings, even within the same gene, mutations can have significantly different functional impacts, highlighting the importance of precise mutation interpretation.

3billion participated in the challenge using ClearVariantPro, an AI model that optimizes target embedding space based on multitask learning. The evaluation compared and validated leading global models, including Google DeepMind's AlphaMissSense and Meta's ESM. ClearVariantPro predicted functional enhancement, functional loss, and changes in drug response with high accuracy.

This evaluation was conducted by Professor Russ Altman's team at Stanford University, and 3billion's model was selected as the best model in the FGFR task. These results demonstrate that 3billion has secured a technological foundation for precisely interpreting genetic mutations, suggesting potential for expansion into the fields of precision medicine and new drug development, as well as for advancing the diagnosis of rare diseases.

Geum Chang-won, CEO of 3billion, said that this achievement is an internationally verified case of the competitiveness of genome interpretation AI technology, and that he plans to continue expanding the application of the technology to mutation-based precision medicine and new drug development in the future.

• See more related articles