Industrial micro­bi­ology is a multi-billion dollar market, and tens of thousands of scientists and experts contin­u­ously work to improve the microbes they use. Micro­bi­ol­o­gists have tradi­tion­ally made better microbes by:

a. selecting and cultivating the best-performing strains,

b. inducing mutations in the hope that some mutations would be helpful, or more recently,

c. manip­u­lating the microbial genome.

Cultivating microbes was the only improvement process available until the 1940s, when scientists jumpstarted the science of ​“strain selection” while scaling penicillin production. The first high-performing strain they selected nearly tripled penicillin secretion from 60 micrograms per milliliter to 150. They exposed later strains to x‑ray and ultraviolet radiation, which eventually led to the Wisconsin family of microbes that manu­fac­tures most penicillin today.

Careful selection and radiation exposure alone prompted a hundredfold increase in Penicillium chrsyogenum’s productive capa­bil­i­ties — an impressive feat. But after decades of repeated use for this and a host of other appli­ca­tions (many involving yeast), selective curation and uncon­trolled mutation became less effective tools.

Scientists began searching for ways to manipulate the bacterial and yeast genomes directly with genetic engineering tools. This process, even with deep genome sequencing, proteomic tools for under­standing what molecules genomes produce, and some improve­ments in automation and lots of computer power, turned out to take the best companies in the world years if not decades and billions of dollars.

The problem is our incomplete knowledge of which microbial genes influence which char­ac­ter­is­tics that can be tuned in microor­gan­isms to produce human useful molecules while leaving the living systems viable and able to reproduce. Unfor­tu­nately, as this influential Zymergen blog post explains, ​“all microbes have significant (25%+) ​‘dark regions’ in their genomes in which the functions of constituent genes are completely unknown.” Finding the right portion of microbial DNA to manipulate is a shot in the dark, and even F1000 companies full of smart folks have met with failure on many key projects.

Our bacterial (and similarly, fungal/yeast) genome map is patchy in part because of its unbe­liev­able size. Conser­v­a­tive estimates of the design space — ​“all the combi­na­tions of all the base pairs in a microbe’s DNA” – are around 4 raised to the power of 3,000,000. This volume of possible combi­na­tions prevents even the most skilled micro­bi­ol­o­gist, or entire teams of them, from knowing how genetic changes will impact bacteria. Even with the increasing pace of genetic research, the sheer size of this space means that much of it will remain uncharted for decades.

The Zymergen Advantage: AI and Robotics for ​“Radical Empricism” and ​“Biore­ach­ables”

Zymergen is side­step­ping this enduring knowledge gap by having algorithms (not scientists) make and test hypotheses about microbe performance. Their AI software sends experiment instruc­tions to an automated robotic platform, which carries out the experiments and sends the results back to the software. The results of past experiments determine which genes are edited in future ones, in a powerful closed loop learning process that is well suited to deep learning innovation. This platform signif­i­cantly reduces human error and carries out experiments more than 1000 times faster than a researcher could manually perform, even a large team of researchers.

Zymergen’s scientists do not have to know in advance how each genetic change influences the microbes. To quote Zymergen CTO Aaron Kimball: ​“We get paid because it works, not because we understand why”. This outcome-oriented approach – delivered on a high-speed, heavily automated, repeatable, and verifiable platform –puts Zymergen years (if not decades) ahead of their competition.

Zymergen was founded in 2013 by Josh Hoffman, Dr. Zach Serber, and Dr. Jed Dean. They had all seen large-scale engineering problems in the course of their work: Josh’s as an elite industry consultant at McKinsey and global investment banks, Zach’s as a synthetic biologist, and Jed’s as an automation engineer in the life sciences. The three of them created Zymergen to solve those problems. They aimed to overcome industrial microor­ganism efforts’ inac­ces­sibly huge design space, large knowledge gaps, and costly drawbacks of human error. Zymergen has solved these microbe engineering problems, and it is enabling massive progress in a related field as well: materials science.

Zymergen is using the microbe genome as a search space for new bioman­u­fac­turable chemicals and materials. They call these molecules ​“biore­ach­ables”, and the databases they’ve built on microbe metage­nomics and biore­ach­ables are the world’s largest. Zymergen is working with materials scientists in industry and academia to search the thousands of new molecules they’ve discovered for the most promising commercial candidates.

Zach Serber, Zymergen’s CSO and co-founder, knows that bioman­u­fac­turing will shape the future of materials science. He explains that bioman­u­fac­turing is better at making certain molecules than petroleum-based processes, which are used in many common materials. For example, biological processes are better at producing:

a) chiral molecules of a particular orientation,

b) molecules made of non-hydrocarbon atoms, and,

c) polymers with diverse chemically reactive groups.

Chiral molecules are asymmetric – their mirror image cannot be transposed on the original. The original and mirror image are called left-hand and right-hand molecules. Chemical manu­fac­turing usually results in an equal distri­b­u­tion of left- and right-hand molecules. Biological manu­fac­turing normally makes only one or the other. This is important because chirality influences how molecules act in a biological setting. Aspartame, the artificial sweeter in Equal and NutraSweet, would taste bitter if you somehow ate the wrong chiral molecule. In a more serious situation, the left-hand molecule of some drugs is beneficial, while the right-hand molecule is poisonous.

Purifying chiral molecules after chemical manu­fac­turing is an expensive process, with purified products sometimes costing a thousand times more than the original mixture. Bioman­u­fac­turing makes purifi­ca­tion unnecessary because it only produces one type of chiral molecule. Chirality is also important in materials science. PLA, a biodegrad­able plastic popular in disposable tableware, has thermal and mechanical properties that change signif­i­cantly with its distri­b­u­tion of chiral molecules. Biological manu­fac­turing makes the selective production of chiral molecules an easier, cheaper process.