AI is transforming how scientists mine biology for therapeutics, allowing researchers to dive deeper into both human and ancient biology to look for novel therapeutic molecules. We spoke with César de la Fuente (left) – Presidential Associate Professor at the University of Pennsylvania (PA, USA) – at ELRIG’s Drug Discovery 2025 (21–22 October; Liverpool, UK) to learn about how he has developed AI models to discover and design antibiotics. César’s hunt for novel antibiotics is spurred by the problem of antimicrobial resistance (AMR), which contributed to approximately five million deaths around the globe in 2019. If left unaddressed, this number is projected to double by 2050. In this feature, we explore the role AI can play in scouring both the human proteome and ancient datasets for potential antibiotic molecules as well as its generative capabilities for designing de novo molecules. AI meets antibiotic discovery and design Identifying new antibiotic candidates for preclinical screening using traditional methods can take many years. Now, researchers can achieve this in just a few hours. So, what’s changed? “Our research has evolved tremendously. Our early work leveraged evolutionary computation – taking inspiration from Charles Darwin – to evolve molecules in such a way that they might become therapeutic. However, in the last 7+ years, we’ve increasingly used deep learning models to accelerate the pace of discovery.” De la Fuente explained that a key shift was treating biology as information: DNA as a four-letter code and proteins as 20-letter sequences. That framing enabled his team to design algorithms…
