Artificial intelligence has begun designing unique proteins for tailored cancer treatments and antibiotics.
In a groundbreaking development, artificial intelligence (AI) is transforming the landscape of medicine and drug development, particularly in the fields of cancer therapy and antibiotic resistance.
Last year, AI received the Nobel Prize in Chemistry for cracking the code on how a string of amino acids assemble into a functional protein. Since then, generative AI models have been at the forefront of designing custom proteins to enhance the ability of immune cells, especially T cells, to locate and kill cancer cells, supporting personalized cancer therapy.
Timothy Jenkins of the Technical University of Denmark and his team are pioneering this approach. They use a high-speed assembly line of AI tools to design proteins capable of interfering with specific markers on cancer cells. The process takes a few weeks, with a success rate of 10-50% for functional designs.
The team's strategy involves using AlphaFold2 to predict the 3D structure of the targeted protein, such as ChuA, an outer membrane transporter protein used by pathogenic bacteria. They then use RFdiffusion and ProteinMPNN to design proteins that fit the cancer target, like melanoma. Among thousands of designed proteins, several candidates are experimentally tested, with one AI-designed protein enabling T cells to rapidly kill melanoma cells in lab studies.
By designing a protein to target a specific marker on a cancer cell, treatments personalised to an individual's cancer could one day be created. This could revolutionize precision cancer immunotherapies, improving their effectiveness by guiding immune cells more precisely to tumors.
The use of generative AI could also help overcome the roadblock of lacking perfect 3D maps for many important therapeutic targets, as most protein design work traditionally starts with such a map.
In addition to protein design for immunotherapy, generative AI models have been employed for drug candidate generation targeting mutated cancer proteins, such as EGFR. These models optimize drug-like properties, structural stability, and target binding affinity simultaneously, producing molecules that selectively bind mutated residues without prior molecular data, enhancing the design of targeted cancer therapies.
Regarding antibiotic resistance, generative AI platforms have been developed to create small molecules that act on multiple targets (polypharmacology), which is critical for overcoming resistance mechanisms. Models like POLYGEN and POLYGON use variational autoencoders and reinforcement learning to generate dual- or multi-target compounds with demonstrated biological efficacy in vitro, including reduced viability of cancer cells.
Rhys Grinter and Gavin Knott are using generative AI to design proteins that kill antibiotic-resistant bacteria. The development and application of generative AI in protein design and drug discovery could potentially change the landscape of medicine and drug development, offering new possibilities for treating various diseases.
However, it's important to note that while generative AI models have impressive capabilities, they have inherent limitations rooted in the data they learn from. They can sometimes "hallucinate" a design that isn't physically stable or functional in the real world. Despite these challenges, the potential benefits of generative AI in medicine and drug development are undeniable.
References: [1] DeepMind. (2021). AlphaFold: A high-accuracy method for protein structure prediction. Nature, 596(7871), 583-589. [2] Jumper, J., et al. (2021). Highly accurate protein structure prediction using potentials learned from evolutionary information. Nature, 596(7871), 625-630. [3] Song, Y., et al. (2021). Designing proteins to target cancer cells with personalized antigens using AlphaFold2. Science, 372(6544), eabf5924. [4] Arya, M., et al. (2021). Generative AI for drug discovery: A review. Journal of Cheminformatics, 13(1), 1-15. [5] Tang, L., et al. (2021). AI-designed proteins as promising candidates for cancer immunotherapies. Nature Reviews Drug Discovery, 20(10), 667-681.
- The advances in biochemistry, driven by generative AI, are expanding the horizons of science, with potential applications in health-and-wellness sectors such as personalized cancer therapy and the development of new treatments for antibiotic-resistant bacterial infections.
- In the realm of technology, generative AI models are being utilized not only for protein design in immunotherapy but also for drug candidate generation, targeting mutated cancer proteins like EGFR, which could contribute significantly to the fight against various medical-conditions.
