VAEBAC: A representation-learning framework for proteome-scale prediction of amyloidogenic proteins and functional stratification of nucleic-acid-binding proteins.

VAEBAC predicts amyloidogenic potential in nucleic-acid-binding proteins across diverse organisms, revealing significant enrichment in transcription regulation functions.

• Why it matters: Understanding how aggregation susceptibility is distributed within regulatory proteomes can clarify mechanisms of genome regulation and disease, addressing gaps in knowledge about protein aggregation in vital cellular processes.
• What they did: Using VAEBAC, a deep learning framework trained on amyloid sequences and annotations, the study performed proteome-wide predictions of amyloidogenicity in proteins from E. coli, B. subtilis, S. cerevisiae, and humans, analyzing their functional roles.
• The result: Predicted amyloidogenic proteins are enriched in transcription-related functions, with spatial separation between aggregation-prone regions and DNA-binding sites, supporting a conserved model of functional stratification of aggregation susceptibility.