Biology is computation. Cells, the smallest units of life, are incredible in many ways. We often think of them as biological machines with symphonies of chemical reactions that enable cells to live, thrive, and divide. But, in many ways, cells can be thought of as supercomputers capable of performing complex calculations that allow them to communicate, grow, survive, and reproduce (among many more distinct and complex decisions that occur across timespans from fractions of seconds to many hours). Individual living cells are capable of general-purpose computation and express the attributes of a Turing Machine. Similar to executable software in the digital realm, cells can execute biological code and copy that code into additional machine “instances” (i.e., progeny cells). Some scientists estimate that the computational power of cells may exceed the predicted limit of Moore’s law by several orders of magnitude.
Over billions of years of evolution, cells banded together to form more complex systems and organisms. If you think hard enough about it, the sophisticated computation and information processing required for multicellularity is truly mind-boggling. Our immune system, for example, is a delicate ecosystem of interactions between many human cell types and commensal microbes resident throughout our body. How the host immune ecosystem was established and built over time, with complex algorithms to detect self from non-self, remains a great mystery. But what is clear is that cells have an incredible fidelity and sophistication with which they can detect even the slightest differences in molecules and move into action with great purpose and intent when things seem awry.
In contrast to animals, plants have different methods for dealing with intruders. Recently, we shared our enthusiasm for plant chemistry and biology. Plants evolved as largely stationary organisms, and given their lack of mobility, they’ve had to get creative with how they protect themselves from pests. Plants fend off predators with an armament of incredible chemical diversity that can repel intruders directly or call in help through chemical attraction mechanisms. For example, an attack by insects induces plants to emit blends of compounds (terpenoids being an important class of molecules here) that both act as repellents and attract other insects that prey upon herbivores.
What is clear to us is that biological systems have evolved to solve problems. They are highly efficient machines capable of responding to complex signals. How do we utilize the advantages presented by biological systems to solve great societal problems? Programming biology can transform fields as wide as medicine, agriculture, energy, and many more.
In creating new medicines, we use the term drug discovery rather than drug design for understandable reasons. Historically biomedical researchers have discovered small molecule drugs through a serendipitous game of brute force. Once we have a target identified, we usually throw a bunch (thousands to millions) of molecules at it to see what sticks. This approach is known as high-throughput screening. We have had the most success working on the canonical drug targets with well-defined and distinct small molecule binding pockets in their active sites. These canonical targets are sometimes referred to as the druggable proteome, though specific definitions of druggable aren’t specific and are constantly changing (e.g., as in kinases).
Recently, AlphaFold by Deepmind has made an enormous leap forward in our ability to predict protein structure from its amino acid sequence alone. Deepmind, collaborators, and other academic and industry groups have followed these achievements with more data, predictions, and unique applications. AlphaFold opens paths to many amazing downstream applications but hasn’t solved drug discovery. The reasons are myriad, but a few worth mentioning: 1) many of the predictions assume a naked protein in isolation, which is never the environment in which a drug needs to work in a patient; 2) the predictions only work for structured regions of proteins, which in some cases aren’t the interesting areas to target; 3) the predictions don’t take into account the multiple conformations that proteins adopt, and don’t express the dynamic motion on cell biology-relevant timescales. These facts do not diminish the significance of AlphaFold. Instead, they highlight the profound complexity of biology and the multi-factorial challenges of drug discovery.
But, as we alluded to, what if we leveraged biological systems to help us solve some of the challenges we face in drug discovery? What if we could program bugs to find hits for challenging, undruggable targets? Today, we’re excited to announce our investment in a team working on exactly that — Think Bioscience.
The platform is a creative step forward in directed evolution. In a simplistic rendering, the Think Bioscience platform programs a hit-finding function into microbes and says, “find a way to inhibit this target or die.” They simultaneously equip the microbes with chemical tools of plants, encode natural product pathways, and let natural selection run. Review the proof of concept led by the team for further details (Sarkar et al.).
The platform’s capabilities are powerful and profound — it recapitulates much of the native cellular environment, integrates relevant chemical diversity, and allows for the production of challenging to synthesize compounds.
By its design, the platform is a functional assay in a live cell environment. This design means the target needs to be inhibited (rather than just bound) to demonstrate an effect. This biocomputation occurs in the crowded cellular environment where the protein is free to flop around and behave as it would in the real world. The dynamic nature allows interrogation of allosteric sites that present themselves on physiologically-relevant timescales inaccessible to most computational approaches. Allostery can be a crucial mechanism, especially in the context of currently undruggable targets that compose, by some estimates, up to 85% of the human proteome. Phosphatases, for example, are a target class with renewed interest as a result of allosteric mechanisms. Phosphatases are historically difficult to drug because the active site is especially polar, meaning drugs that bind tightly will be strongly hydrophilic, hurting other important druglike property requirements.
Additionally, the platform enables significant chemical diversity that is important for engaging dynamic regions of proteins. Many large screening libraries are biased toward historical drug targets and active sites of those targets. However, natural product pathways (such as terpenoids, alkaloids, and phenols) provide a wide array of chemical diversity. These have been an incredibly rich resource for a long time in the form of ancient, natural medicines and modern pharma drug development; more famous examples include paclitaxel, artemisinin, and digitalis. As a bonus, the bug is also the manufacturing engine for these compounds that eliminates the need to source plant extracts from all over the world or develop long, tricky total synthesis pathways.
Not only has Think Bioscience built an elegant platform, but the team also brings together experts at the intersection of synthetic biology and drug discovery. Jerome Fox, Philip Jeng, and Matt Traylor bring a wealth of expertise to this unique intersection of technologies. We are thrilled by the opportunity to partner with the Think Bioscience team and look forward to their continued progress.
Think Bioscience represents a unique example of the Super Evolution applied to drug discovery. A platform with an advanced engineering approach that aims to solve undruggability problems across diverse target classes. While we’re far from automating medicinal chemistry, combining bugs, biosynthesis, and natural product chemistry represents an exciting path forward.
