PuSH - Publikationsserver des Helmholtz Zentrums München

Olayo-Alarcon, R. ; Pugno, D. ; Müller, C.L.

Protein Content-Based Microbial Representations Improve Predictions of Antimicrobial Activity.

In: (34th International Conference on Artificial Neural Networks-ICANN-Annual, 9-12 September 2025, Kaunas, LITHUANIA). Berlin [u.a.]: Springer, 2026. 235-237 (Lect. Notes Comput. Sc. ; 16072)
Deep learning-assisted strategies have enabled the identification of novel antimicrobial compounds with broad-spectrum activity. While these discoveries hold great value, targeted therapies that are effective against specific pathogens and have minimal impact on commensal microbes remain a key objective in the fight against antibiotic resistance. Here, we address this challenge by developing and evaluating proteome-based representations of bacterial species. We find that these representations enable species-resolved predictions of a compound's growth-inhibiting activity. In particular, microbial representations based on the presence of homologous protein groups enable generalizable predictions for unseen compound-microbe pairs. These microbe-specific predictions more accurately recapitulate findings from large-scale chemical screens involving compounds and bacterial pathogens not encountered during model training. We envision that this framework can facilitate the development of targeted therapies.
Weitere Metriken?
[➜Einloggen]
Zusatzinfos bearbeiten [➜Einloggen]
Publikationstyp Artikel: Konferenzbeitrag
Schlagwörter Microbe representations; Antimicrobials; Proteome
ISSN (print) / ISBN 0302-9743
e-ISSN 1611-3349
Konferenztitel 34th International Conference on Artificial Neural Networks-ICANN-Annual
Konferzenzdatum 9-12 September 2025
Konferenzort Kaunas, LITHUANIA
Zeitschrift Lecture Notes in Computer Science
Quellenangaben Band: 16072, Heft: , Seiten: 235-237 Artikelnummer: , Supplement: ,
Verlag Springer
Verlagsort Berlin [u.a.]
Institut(e) Institute of Computational Biology (ICB)