Can deep learning prevent a return to the pre-antibiotic era?
In short: As bacteria evolve to bypass our strongest medicines, the pharmaceutical industry has largely retreated from antibiotic research. But a new 'lab-in-the-loop' approach is using artificial intelligence to scan billions of molecules, aiming to discover entirely new classes of drugs before the crisis becomes unstoppable.
The world is facing a mounting antimicrobial resistance (AMR) crisis. Driven by the evolution of superbugs and the inappropriate use of antibiotics in medicine and agriculture, the United Nations estimates that by 2050, antibiotic resistance could cause 10 million deaths annually, surpassing the impact of cancer. This threat is compounded by an economic reality: pharmaceutical companies have largely abandoned the space because the return on investment is too low, leaving a dwindling discovery pipeline. The scale of this burden is immense; in the US and Europe alone, these resistant infections add 11 million hospitalization days and over $20 billion in annual healthcare costs. Currently, antibiotic resistance is already responsible for approximately 700,000 deaths every year worldwide.
To combat this, the Antibiotics-AI Project at MIT, led by bioengineer Jim Collins, is integrating deep learning with experimental biology. Their method employs a 'closed loop' or 'lab-in-the-loop' system: researchers empirically test thousands of compounds against priority pathogens, and the resulting data is used to train deep neural networks. These models then predict antibacterial activity from molecular structures, allowing scientists to search through over 70 billion molecules in silico or even design brand-new molecules from scratch. This computational screening has already been used to empirically test over 38,000 molecules against 3 bacterial pathogens and over 100,000 molecules against 3 different human cell lines. This massive scale allows for the identification of non-toxic candidates through de novo generation.
The results are already emerging. By employing deep learning, scientists identified potent molecules for Neisseria gonorrhoeae, while explainable graph neural networks revealed a fresh antibiotic category capable of fighting MRSA and vancomycin-resistant enterococci. The initiative aims to find seven distinct antibiotic groups within seven years, focusing on the most lethal microbes expected to threaten human wellbeing throughout the coming thirty years. The project's target list includes dangerous pathogens such as Mycobacterium tuberculosis, Klebsiella pneumoniae, and Staphylococcus aureus. This research follows a long history of antibiotic development that began with milestones like Salvarsan in 1910 and Penicillin in 1928.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: Jim Collins: How we're using AI to discover new antibiotics
- Also: Antibiotics-AI Project — Collins Lab, MIT
- Also: Artificial Intelligence and Antibiotic Discovery
To cite this card, link to https://intellectplus.co.uk/app/technology-how-we-re-using-ai-to-discover-new-antibiotics.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
