It’s 2022, and we still can’t predict what a molecule does based off of its looks.
Plants, bacteria, and other living systems are really good at producing and evolving libraries of molecules for doing all sorts of things – insect repellant, sending a warning, fighting infection, and more. And often, those small molecules benefit us as well. 34% of drugs approved by the FDA between 1981 and 2010 come from natural sources. That should come as no surprise, considering the antioxidants in aloe vera that treat sunburn so effectively, or the miracle drug artemisinin from sweet wormwood.
But – newsflash – there are still plenty of diseases that we still suffer from, and we are still in need of drugs that can fight them.
Computational scientists would like to be able to match chemical structure to function, and predict what molecule we need based off of what we want it to do. For example, if we want to fight cancer with a molecule that inhibits mitosis in human cells, we should run a search on known molecules that do that kind of thing, get a rough estimate of what they look like, and predict some new thing that looks typical enough to do just that. Unfortunately, these methods tend to predict impossible molecules. That is, the small molecules they suggest are too chemically complicated for synthesis in a chemistry lab or even by a cell.
Chemists would like to be able to study the thermodynamics of protein-ligand interactions to predict what kinds of molecules they need. With this approach, we consider that many drugs act by binding to the source of the problem. Sticking with the cancer example, there are many kinases involved in human cell mitosis. Small molecules that inhibit kinases, then, could fight cancer. That’s exactly what a lot of clinical kinase drugs do. To be a good inhibitor, a small molecule must bind tightly to that kinase, and should also bind specifically to that kinase (considering that there are lots and lots of different kinds of kinases, you don’t want to go about inhibiting all of them).
A molecule binding to an enzyme is actually a kind of chemical reaction, so it is governed by a range of thermodynamic properties – what is the enthalpy of this reaction? What is the entropy of the reaction? These kinds of properties are very difficult to measure and predict when designing small molecules.
Then, we could also go the natural product route. Nature produces molecules all the time, many of which we can harvest and convert to drugs for ourselves. But nature rarely produces these molecules at rates fast enough for us. That’s ok, though, right? Once nature has discovered the structure of the molecule we need, we can just make it ourselves, right? (We do this with vanilla often enough.) Not so fast – biology is really good at making molecules, but us? Our molecule-synthesizing abilities are more, “meh.” In a chemistry lab, we can dump extraordinary amounts of money into making “natural products” on our own. This is why artemisinin, the miracle drug for malaria, is so expensive. It’s just so difficult for us to make, and we’re not getting the world’s demand for it from wormwood anytime soon.
But wait a minute – plants and microbes go about designing new molecules all the time. Think about it: A tree is surrounded by fellow trees. It is suddenly attacked by a leaf-eating bacterial infection (or something). It releases a chemical into the air (or root system) to warn its neighbors, and it releases a chemical to fight the infection. The bacteria respond by releasing a chemical of their own to protect themselves from this defensive strategy. The tree mutates its pathway for small molecule design to come up with a bigger, better signal and a bigger, better antimicrobial. An evolutionary arms race has begun, and it actually happens all the time.
Living systems are not only a source of natural products – they are a source of synthesis, discovery, and design!
What if we could harness the power of living systems to design small molecules for ourselves?
This is what a new paper proposes we do. Microbially Guided Discovery and Biosynthesis of Biologically Active Natural Products from Sarkar et al. (2021) is a brand-new research paper that harnesses the power of biology to design and screen thousands of small molecules at once. First, you must set up a cell with a sort of genetic logic gate – something that pressures it to evolve the way we want it. Then you give it a library. In Microbially Guided Discovery, the cells are given a library of terpene synthases. Terpenoids are a diverse set of natural products, and a favorite tool for plants and microbes engaged in an evolutionary arms race. You can mutate enzymes in this pathway to diversify the terpenoid library even further and use your engineering microbes to find the right small molecules for you.
Recall the kinases? They are complemented by phosphatases, a type of enzyme that removes phosphate groups.
Phosphatases are also heavily involved in cancerous cell pathways. But unlike kinases, old methods for small molecule design have come up short. There are no FDA-approved drugs for phosphatase inhibition on the market right now, despite increasing evidence that that kind of thing could be very useful against some types of cancer, and other major diseases like Alzheimer’s and diabetes. These much-needed drugs are exactly what is sought in the Microbially Guided Discovery paper: cell survival is linked to whether the cell can produce small molecules that inhibit a phosphatase.
This new method seems an exciting new way to screen and select small molecules for us. We don’t need to predict what our molecule will look like, or engineer our molecule to bind tightly, and those things are almost too hard for us to do now anyway. Instead, we set a cell up to do all that for us. In the paper, they noted a few potential drug targets. Will any make it to clinical trial? Can the method be tweaked to search more specific or more diverse types of molecules? Time will tell, and the science is promising.