Use of Artificial Intelligence in the Design of Small Peptide Antibiotics Effective against a Broad Spectrum of Highly Antibiotic-Resistant Superbugs Artem Cherkasov ‡,¶ , Kai Hilpert †,¶ , Håvard Jenssen † , Christopher D. Fjell ‡ , Matt Waldbrook † , Sarah C. Mullaly † , Rudolf Volkmer § , and Robert E.W. Hancock †, * † Centre for Microbial Diseases and Immunity Research, University of British Columbia, 2259 Lower Mall Research Station, Vancouver, British Columbia V6T 1Z3, Canada, ‡ Division of Infectious Diseases, Faculty of Medicine, University of British Columbia, 2733 Heather Street, Vancouver, British Columbia V5Z 3J5, Canada, and § Institut fu ¨r Medizinische Immunologie, Universita ¨tsklinikum Charite ´ , Humboldt-Universita ¨t zu Berlin, Schumannstr. 20-21, 10117 Berlin, Germany. ¶ These authors contributed equally to this work. T he spread of antibiotic resistance among bacte- rial pathogens especially in hospital environ- ments but also now in the community has oc- curred at an alarming rate ( 1, 2). There is now a proliferation of so-called “Superbugs” that are resistant to multiple or all antibiotics and severely limit treatment options; these include methicillin-resistant Staphylococ- cus aureus (MRSA), vancomycin-resistant enterococci (VRE), and multidrug-resistant Pseudomonas, among others, and cause hundreds of thousands of infections annually. Despite these alarming trends and the obvi- ous need for new interventions, pharmaceutical compa- nies have largely withdrawn from the field of anti- infectives (3), and only two structurally novel antibiot- ics have entered the market in the past 40 years (4). Virtually all species of life produce cationic antimicro- bial (host defense) peptides. These peptides can be di- rectly antimicrobial and/or can play an important role in the functioning and orchestration of innate immune and inflammatory responses of mammals, amphibians, and insects (5-7). These peptides are typically 12-50 amino acids in length with 2-9 excess basic residues (arginine or lysine) and up to 50% hydrophobic amino acids; they fall into four major structural categories based on their amphiphilic conformations that are pre- formed or occur after membrane interaction, namely, -structures with 2-4 -strands, amphipathic -helices, loop structures, and extended structures (5). *Corresponding author, bob@cmdr.ubc.ca. Received for review October 2, 2008 and accepted November 16, 2008. Published online December 4, 2008 10.1021/cb800240j CCC: $40.75 © 2009 American Chemical Society ABSTRACT Increased multiple antibiotic resistance in the face of declining anti- biotic discovery is one of society’s most pressing health issues. Antimicrobial pep- tides represent a promising new class of antibiotics. Here we ask whether it is pos- sible to make small broad spectrum peptides employing minimal assumptions, by capitalizing on accumulating chemical biology information. Using peptide array technology, two large random 9-amino-acid peptide libraries were iteratively cre- ated using the amino acid composition of the most active peptides. The resultant data was used together with Artificial Neural Networks, a powerful machine learn- ing technique, to create quantitative in silico models of antibiotic activity. On the basis of random testing, these models proved remarkably effective in predicting the activity of 100,000 virtual peptides. The best peptides, representing the top quartile of predicted activities, were effective against a broad array of multidrug- resistant “Superbugs” with activities that were equal to or better than four highly used conventional antibiotics, more effective than the most advanced clinical can- didate antimicrobial peptide, and protective against Staphylococcus aureus infec- tions in animal models. A RTICLE www.acschemicalbiology.org VOL.4 NO.1 • ACS CHEMICAL BIOLOGY 65