This week I read Synthetic Biology of Natural Products Engineering: Recent Advances Across the Discover−Design−Build−Test−Learn Cycle.
I originally chose to study biological engineering because of the spider silk studies being done at Utah State University (USU) as part of the International Genetically Engineered Machine (iGEM) competition. Students at USU had engineered E. coli to produce spider silk proteins. They also later engineered those spider silk genes into a goat which was able to produce spider silk proteins in its milk. As a lifelong science lover I had never been more fascinated by a set of experiments.
In my recent job search I’ve been reminded why I chose my undergraduate major. My small company/startup background has led me to research synthetic biology startups, many of which are using synthetic biology and genetic engineering to produce natural products (NPs) in a lab setting. I’ve been struck at just how much work is still being done to engineer organisms to produce NPs that are otherwise difficult or impossible to produce (e.g. farming spider silk).
This week I looked for articles centered on this type of work and landed on this article by Foldi et al. Although I’ve done a good bit of synthetic biology work myself, I’ve typically worked with well-characterized genes and model organisms. This article detailed recent refinements to the genetic engineering process, from gene/metabolic pathway discovery through analysis of final compounds produced.
Foldi et al. begin by describing genome mining, the process by which biosynthetic gene clusters (BGCs) for a specific NP are identified. It’s no surprise that machine learning (ML) is now at the forefront of these efforts. The article even outlines many of the competing ML software tools for genome mining, comparing the strengths and weaknesses of some of the more prominent tools.
Next the article describes how biosynthetic strategies are designed. Since genes responsible for biosynthesis tend to be found in genomic clusters, scientists can find and study various enzymes involved in a synthesis pathway of interest. This also means that, “when a desired compound cannot be produced by an existing natural pathway, tools for retrobiosynthesis can be used to design a pathway.” There is specialized software to predict how desired pathways can be constructed. As an undergraduate I studied Constraint-based reconstruction and analysis (COBRA) to predict feasibility of sustaining target metabolite production, though it’s clear this type of technology has come a long way in the last few years.
Advances in genetic engineering tools (such as CRISPR-Cas9) allow labs to produce NPs that could not typically be produced in labs. The article gives two examples. First, scientists can engineer related organisms to produce NPs that are generally only produced in non-lab conditions (e.g. deserts or wastewater treatment plants). This would be like engineering one strain of Streptomyces to express a BGC from another type of Streptomyces. Second, labs can take advantage of “cross-kingdom heterologous hosts”. For example, certain types of yeast have been successfully engineered to express plant BGCs. Overall, genetic engineering advances allow scientists to build ideal systems, whether that means selecting a strain that has greater capacity to accumulate a target compound or moving a biosynthesis pathway to an entirely different branch of the taxonomic tree. Advances in lab techniques also allow for “coculturing”, in which multiple organisms are cultured together and are each responsible for specific portions of metabolite production. There are also studies investigating the potential of cell-free systems to produce products that would either be too burdensome or too toxic for cells to produce.
There are also a host of challenges associated with characterizing produced compounds and confirming a product is the expected compound. Ongoing efforts seek to improve spectrometry and software/ML tools to address these issues. ML models are also being used to analyze and learn from synthetic biology datasets.
Personally, I’m eager to get up to speed on current NP production tools. I’m working through a couple ML courses, after which I’d love to dive into metabolic modeling tools. It looks like in silico is the way of the future.
The full article can be accessed here through ACS Synthetic Biology.
*This post is part of an ongoing series of informal article summaries, intended to keep me up-to-date on industry trends and the latest research. I identify articles through an RSS feed I curated, containing articles from ACS Synthetic Biology, Cell, Nature, and ProQuest (Synthetic Biology).
**Headline image taken from iStock