computational foraging across the proteomic frontier
Nature is a process for generating locally negentropic solutions to the challenge of universal thermodynamic destiny. That's a big challenge.
On Earth, Nature has developed molecular machines as its keystone technology. These molecular machines are proteins. Their blueprints are written as DNA inside organismal genomes. Reading, understanding, and modifying these blueprints is the field of proteomics.
Entomolecular uses cutting-edge computational tools to look for novel protein configurations in the growing longtail of sequenced insect genomes.
These genomes are already public, and more of them are released every year. But the vast majority have never been searched for short amino acid sequences that could be useful to humanity. In that tail the sequences have drifted. The folds have not.
Insects are the most abundant and diverse terrestrial animals on the planet. Humanity has named about 1 million species, roughly two-thirds of every animal species ever described, but we believe that is less than 20% of the true total. Their diverse biologies and behaviors means genetic diversity which means protein diversity.
Yet less than 10 thousand insect genomes have been sequenced, and only a tiny fraction of those have been annotated such that their proteomes are legible to us. For the majority of sequenced insect genomes, let alone all insect genomes, their proteomes remain largely inscrutable. This is an enormous opportunity.
Parasitoid wasps use complex neurotoxins to control the behavior of hosts. Ants farm massive fungus gardens 5 meters underground. Mosquitoes take blood and the cover of anesthetizing saliva. Other insects spin silk, or force a plant to grow a gall, or survive high altitudes and freezing temperatures. Each of these unique lifestyles is made possible by a set of proteins.
This is why Entomolecular is focused on this nearly untapped natural resource.
Classical computers, like proteins, are wonderful machines. But when it comes to designing and folding proteins, they face a gargantuan hurdle. For example, a single small chain of 100 amino acids, with one of the 20 possible residues at each position, contains 20100 possible configurations. A game of chess, the classic combinatorial example, and whose full solution still eludes our most powerful computers, has about 10120 possible paths — our single 100 amino acid chain is more complex by an order of more than 10 billion.
Once you decide on a sequence of amino acids, folding multiplies the problem. If the chain has just 3 degrees of freedom at each of its 100 links, it can take on 3100 possible shapes. If a computer were to try one shape every trillionth of a second, all the folds would last 1018 ages of the universe.
Meanwhile, Nature has been searching this vast computational space for billions of years, producing a refined library of incredible machines, vetted by adversarial evolutionary ecology, and waiting to be discovered by curious humans. This is why searching this naturally prepared library is still our best method of developing useful proteins for use in medicine, agriculture, and industry.
Given a protein or target molecule of interest, we use computational search and modeling through to narrow the insect proteome down to viable candidate proteins. From there, wet lab assays with real proteins are the crucial next step. Our first search has been for Cys-rich Kunitz class protease inhibitors. This is an overview of our process:
The object of our first search effort is the Kunitz class of protease inhibitors. Secreted versions of the fold turned up several in Hymenoptera species — ants and solitary bees. The matches were structurally homologous to reference protease inhibitors but their sequence identities were only 30 to 40 percent. The shape similarity is the filter, the sequence distance is what makes these findings compelling to continue.
The next step is a protease assay.
Example of a candidate protein
A computed model of a short secreted protein. The shape is a Kunitz domain, and along the structural alignment it shares about 40 percent of its sequence with the reference. The yellow spheres mark the cysteines that are the marquee feature of the disulfide backbone of this class of proteins.
Entomolecular is a bioprospecting startup based in California, started by a small group of ento- and molecular biologists.