AI Is Revolutionizing Medicine: What’s Next?
▼ Summary
– Traditional drug development is a slow process, taking 10-15 years from discovery to approval, which creates a long wait for patients.
– Amgen is integrating AI and machine learning into its R&D to accelerate drug discovery, particularly for serious illnesses like cancer and cardiovascular disease.
– The company’s approach is grounded in large-scale human genetic data, which helps identify genetically validated targets for more successful medicines.
– AI tools at Amgen, such as AMPLIFY and EVOLVEpro, are used to design and optimize novel protein-based drug candidates, significantly speeding up protein engineering.
– The ultimate goal is to scale these AI/ML technologies across the entire drug pipeline to treat a vast majority of disease targets that currently lack effective medicines.
The journey from a scientific concept to a medicine in a patient’s hand is notoriously long, often spanning over a decade. This lengthy timeline presents a significant challenge for both individuals awaiting new treatments and the researchers working to develop them. Companies like Amgen are now integrating artificial intelligence and machine learning into the very core of drug discovery, aiming to dramatically shorten this process and unlock new therapeutic possibilities for serious conditions such as cancer and cardiovascular disease. By leveraging these advanced technologies, the goal is to transform a traditionally slow, trial-and-error maze into a more precise and accelerated pathway to innovation.
At the heart of this transformation is a profound belief in the power of human genetics. Access to vast and diverse genetic datasets provides a critical foundation for identifying targets that have a higher probability of leading to successful drugs. Amgen’s research site in Iceland utilizes one of the world’s richest repositories of human genetic information, including millions of genotypes and hundreds of thousands of whole-genome sequences. This genetic grounding allows scientists to pursue new hypotheses with greater confidence, translating complex biological insights into potential medicines with a renewed sense of urgency. Robust governance frameworks ensure this powerful data is handled with strict security and privacy protections.
The impact is already visible in the laboratory. Nearly every molecule in Amgen’s early research pipeline is now influenced by an AI-driven approach, particularly in the intricate field of protein engineering. After a target is identified, generative AI tools assist researchers in designing novel protein structures with desired drug-like properties, all before any physical experimentation begins. Predictive machine learning models then forecast how these designed proteins will behave in the body. One internally developed model, for instance, can predict a key characteristic for injectable drugs, protein viscosity, with over 80% accuracy using only the amino acid sequence as input. This tight integration between custom-built AI systems and the traditional stages of drug development creates a more cohesive and efficient research workflow.
This acceleration relies on specialized tools built for biological challenges. A collaboration with the Mila AI institute in Canada produced AMPLIFY, an open-source protein language model that deciphers the complex “grammar” of proteins, mapping sequences to their structure and function. It helps scientists understand the biological “sentences” written in the genome to design better proteins from the outset. Complementing this is EVOLVEpro, a system that goes beyond analyzing natural evolution to actively design improvements for medicine. Unlike models trained solely on existing protein data, EVOLVEpro proposes intelligent mutations, learns from experimental results, and uses that new data to refine its future predictions. In testing, it has increased protein activity by up to 100-fold for various therapeutics in just a few experimental cycles, creating a powerful, self-improving loop where AI actually guides scientists toward better experiments.
Looking to the future, the ambition is to apply these AI and machine learning technologies across the entire drug development continuum, from the earliest discovery phases to regulatory submissions. These tools also hold promise for optimizing existing medications to benefit more patients. The ultimate objective is to tackle the vast majority of disease targets that have remained out of reach for conventional medicine. This represents a monumental shift, moving beyond simply increasing speed to fundamentally reimagining what is scientifically possible. The mission remains fixed on serving patients, but the methods are evolving, representing a new chapter in fusing deep biological understanding with cutting-edge technology to confront the world’s most difficult diseases.
(Source: wired)
