By leveraging advanced AI techniques to comb through thousands of recent biomedical abstracts, a collaborative team of South Korean researchers has highlighted two previously overlooked proteins—FGF1 and HGF—as promising targets for conditions like macular degeneration.

Ocular neovascularization, characterized by the abnormal growth of blood vessels in the eye, is a primary driver behind vision-threatening conditions such as age-related macular degeneration. While existing treatments typically focus on inhibiting Vascular Endothelial Growth Factor-A (VEGF-A), many patients experience inadequate responses or resistance over time.

To address this therapeutic gap, a joint research team spearheaded by Professor Yong-seok Moon from Hallym University Kangnam Sacred Heart Hospital, in collaboration with researchers from Seoul National University Hospital and Seoul National University Bundang Hospital, turned to artificial intelligence.

The investigators utilized Graph-based Retrieval-Augmented Generation (GraphRAG) to process 5,562 PubMed abstracts focused on angiogenesis published from 2022 to 2024. This computational approach generated an expansive biomedical knowledge graph comprising 17,842 distinct concepts and 9,555 interconnected relationships.

By mapping biological connections radiating from the established VEGF-A target, the AI-driven analysis successfully spotlighted two alternative proteins: Fibroblast Growth Factor 1 (FGF1) and Hepatocyte Growth Factor (HGF).

FGF1 plays a critical role in cellular proliferation and the genesis of new blood vessels.
HGF regulates cell motility and growth, steering angiogenesis via the c-MET receptor while showing potential links to patients who fail to respond to standard anti-VEGF regimens.
A cross-examination against three major recent literature reviews revealed that neither FGF1 nor HGF had previously been recognized as primary therapeutic targets for ocular neovascular conditions, underscoring the novelty of the discovery. Furthermore, subsequent deep learning-based structural modeling indicated that both proteins feature viable binding pockets capable of accommodating pharmaceutical compounds.

"This breakthrough allows us to explore viable options that have largely flown under the radar in traditional research," noted Professor Moon. "Our next phase involves rigorous follow-up validation to pave the way for alternative interventions for patients unresponsive to current standard-of-care treatments."

The details of this study were officially released in the July 2026 edition of the peer-reviewed international journal Briefings in Bioinformatics.

[Copyright (c) Global Economic Times. All Rights Reserved.]