Antimicrobial resistance is no longer a distant threat; it is a current crisis claiming roughly five million lives annually. By 2050, that number could double. Yet we haven’t discovered a new class of antibiotics in half a century. Traditional drug discovery is slow, expensive, and increasingly ineffective at modifying existing chemical structures. César de la Fuente, a bioengineer leading a cross-disciplinary lab, argues that biology is fundamentally an information system. His team treats DNA nucleotides and amino acids like an alphabet, using deep learning to decipher the organizing principles of life.
This shift in perspective changes everything. Instead of physically sampling soil or plants, de la Fuente’s lab scans digital genomes from living and extinct organisms. Their AI models identify patterns in biological sequences that signal antimicrobial potential. What used to take years of manual screening now takes hours. The algorithms sift through vast datasets to find needles in haystacks, prioritizing a manageable list of candidates for real-world testing. It is a dramatic acceleration of the initial discovery phase, though it does not replace the rigorous lab work that follows.
The lab doesn’t just rely on custom-built models. They integrate tools like ChatGPT and Codex to bridge gaps between disciplines. Biologists who aren’t strong coders can write scripts, while programmers can grasp complex biological concepts faster. This interoperability lowers barriers within the team, allowing for quicker hypothesis generation and data analysis. De la Fuente notes that this transdisciplinary approach is essential because only a tiny fraction of genomic data has a clearly understood function. AI helps connect dots across fields that traditionally operate in silos.
However, algorithmic prediction is only the starting line. A molecule identified by AI must still kill microbes in petri dishes, prove safe for human cells, and pass toxicology screens. Manufacturing scalability and regulatory approval remain massive hurdles. De la Fuente emphasizes that ground-truth experiments are non-negotiable. AI can scratch the surface of biology’s complexity, but wet-lab validation remains the ultimate arbiter of success. The technology speeds up the search, but nature still dictates what works.
