As we venture beyond Earth into the vast unknown of space, one thing is certain: we will encounter life forms and phenomena that defy our current understanding. But here's where it gets mind-bending – what if the building blocks of life on other planets are entirely different from what we know? What if their genetic code doesn't play by our rules? This is the intriguing question at the heart of astrobiology, and it's one that scientists are grappling with as we prepare to explore new worlds.
On Earth, life's genetic code is written in a language of four standard nucleotides, a system that has likely been in place for eons. However, this is the part most people miss: the genomics of extraterrestrial life could be vastly different, employing an entirely new genetic alphabet. Imagine a world where the very essence of life is constructed from a different set of letters, a code we've yet to decipher.
In our quest to understand and manipulate life, scientists are pushing the boundaries of genomics. Through the Artificially Expanded Genetic Information Systems (AEGIS), researchers have demonstrated that non-standard nucleotides can indeed pair up, opening the door to a world of possibilities. While the functionality of these new sequences remains uncertain, it offers a glimpse into the versatility of genetic codes and their potential variations across the universe.
In a groundbreaking experiment, scientists synthesized and tested approximately 300 phage genomes in dishes containing E. coli, resulting in 16 functional phages. These phages, or bacteria-infecting viruses, were designed using 'Evo,' a generative AI model trained on an extensive database of living organisms' genomes. Evo's training data comprised a staggering 9 trillion letters of DNA, encompassing all domains of life, akin to the vast corpora used in large language models.
The complexity of life's functions often arises from intricate interactions within entire genomes, rather than individual genes. Genome language models have emerged as a powerful tool for designing biological systems, but their capacity to generate functional whole-genome sequences has been largely untested – until now.
Here's the controversial part: What if we can not only understand but also create life forms with entirely new genetic codes? In this study, researchers report the first successful generative design of viable bacteriophage genomes using advanced genome language models, Evo 1 and Evo 2. By leveraging these models, they created whole-genome sequences with realistic genetic architectures and specific host preferences, using the lytic phage ΦX174 as a template.
The results were astonishing. Experimental testing yielded 16 viable phages with significant evolutionary novelty. Cryo-electron microscopy revealed that one of these phages employed a DNA packaging protein from a distant evolutionary branch within its capsid. Furthermore, several phages exhibited higher fitness than ΦX174 in growth competitions and lysis kinetics. A combination of these generated phages effectively overcame ΦX174-resistance in three E. coli strains, showcasing the potential of this approach in designing phage therapies against rapidly evolving bacterial pathogens.
This research not only provides a blueprint for creating diverse synthetic bacteriophages but also lays the groundwork for the generative design of useful living systems at the genome scale. It raises thought-provoking questions: Are we playing God by creating new life forms? Or are we simply unlocking the secrets of the universe, one genetic code at a time? What are the ethical implications of such powerful technology? We invite you to share your thoughts and join the discussion – do the benefits of this research outweigh the potential risks, or are we treading into dangerous territory?
Food for thought: As we continue to explore the cosmos and unravel the mysteries of life, are we prepared for the unexpected? And more importantly, are we ready to redefine our understanding of life itself? (Generative design of novel bacteriophages with genome language models, biorxiv.org, open access)
