#Ai-Discovery

2 stories

Research April 30, 2026 · View digest

UTHealth Houston BLMP6 Peptide Selectively Binds Fibulin-4 in Metastatic Triple-Negative Breast Cancer

Mikhail Kolonin's group at UTHealth Houston published preclinical data showing BLMP6, a peptide identified through AI-guided modeling, selectively binds fibulin-4 — a protein highly expressed on metastatic triple-negative breast cancer cells — and not on noninvasive breast cancer or normal breast tissue. A BLMP6 conjugate carrying monomethyl auristatin E suppressed metastasis and improved survival in mouse models, and BLMP6-based fluorescent imaging probes successfully detected metastatic lesions. The paper, published in Molecular Therapy Oncology, identifies fibulin-4 as a new theranostic target.

Research April 30, 2026 · View digest

Nature Communications: Few-Shot AI Pipeline Designs Antimicrobial Peptides Against Carbapenem-Resistant Acinetobacter baumannii

A Nature Communications paper describes a deep-learning pipeline that uses pre-trained protein language models combined with few-shot fine-tuning to identify antimicrobial peptides effective against Acinetobacter baumannii, a WHO critical-priority pathogen. The classification, ranking, and regression modules collaboratively prioritize candidates with high predicted activity, expanding the chemical space accessible to data-poor AMR targets. Lead candidates showed potent in vitro activity against carbapenem-resistant clinical isolates.