Deciphering genetic circuits can be tedious and extremely time-consuming. Modifying or designing genetic circuits from previously identified pathways presents additional challenges.
“There are many possible designs for a particular function, and finding the right design is like looking for a needle in a haystack,” said Dr. Caleb Bashour, a Rice University scientist and lead author of a new study that establishes a new technique aimed at helping researchers discover useful genetic circuits, or DNA designs, much faster than before.
This research nature In a paper titled “Ultra-high-throughput mapping of the genetic design space”.
“We have developed new techniques that allow us to design hundreds of thousands to millions of DNA at once, more than ever before,” Bashour said.
Researchers in the Bashour lab, including co-lead authors Dr. Kshithi Rai and Dr. Ronan O’Connell, then graduate students, and interdisciplinary collaborators including physicists and computer scientists, developed a new technique called CLASSIC, which combines long-range and short-range sequencing to probe genetic complexity. At CLASSIC, the team used AI and machine learning (ML) to design these circuits.
“Our work is the first demonstration that AI can be used to design these circuits,” Bashor said.
Using both long-read and short-read next-generation sequencing (NGS), the research team was able to create detailed maps of long DNA sequences. Long-read sequencing generated large amounts of DNA data, but the process was slow and the results could be noisy. Meanwhile, the simultaneous short read sequence reduced noise and made the sequence clearer.
“Most people use one or the other, but we found that using both together frees up the ability to build and test libraries,” O’Connell says.
“We have invented a way to do this in large batches, which allows us to create very large sets of circuits (known as ‘libraries’),” Rai added.
The research team created a library of proof-of-concept genetic circuits that incorporated fluorescent reporter genes into their sequences. They created a complete sequence of the circuit and tagged it with a short DNA barcode.
These genetic circuits were inserted into human embryonic kidney cells. We then analyzed the fluorescence levels of these cells. Individual cells were classified by fluorescence brightness and tested with short-read NGS to scan for DNA barcodes. This created a master map linking genotype and reporter expression phenotype.
“Eventually, you’ll be measuring many possible designs, but not all, and that’s where building ML models comes in,” O’Connell says.
This data was used to train a model that made predictions not found in the original dataset. O’Connell explained that these predictions were subsequently verified in follow-up experiments. “We have all these predictions. Let’s see if they’re right.”
Follow-up experiments and manual checks on a small random dataset suggest that CLASSIC works well.
“We started lining them up, and first one worked, then the other… and they just started hitting,” Lai said. “All 40 pieces matched perfectly. That’s when we knew we had something.”
“This was the first time we could use AI/ML to analyze circuits and accurately predict untested circuits, because up until this point, no one had been able to build a library as large as ours,” he continued.
CLASSIC is able to analyze large datasets, develop detailed circuits, and accurately predict native genetic circuits, suggesting that this model is useful for developing new genetic circuit designs. Using CLASSIC, the team realized that circuits are variable and have multiple paths to the same result.
“It’s like a navigation app: There are multiple routes to get to your destination, some highways, some back roads, but they all get you there,” O’Connell says.
This technique is suitable for use in combination with high-throughput circuit characterization and AI-driven development in the fields of synthetic biology and biotechnology.
“We believe AI/ML-driven design is the future of synthetic biology,” said Bashor. “As we collect more data using CLASSIC, we will be able to train more complex models to predict how to design even more sophisticated and useful cell biotechnologies.”
Dr. James Collins of the Massachusetts Institute of Technology, an early researcher and founder of synthetic biology, agrees. “Twenty-five years ago, these early circuits showed that living cells could be programmed, but they were built one at a time, each requiring months of adjustment.” Although he was not involved in the research, he commented that the work “represented a transformative leap forward.”
“CLASSIC brings high-throughput engineering to genetic circuit design, enabling exploration of previously inaccessible combinatorial spaces. Their platform not only accelerates the design-build-test-learn cycle, but also redefines its scale, ushering in a new era of data-driven synthetic biology,” he said.
“Synthetic biologists have dreamed of programming cells by piecing together biological circuits from interacting genes and proteins,” said Dr. Michael Elowitz, another prominent synthetic biology researcher and 2007 MacArthur Fellow.
He noted that the field is challenged with understanding and developing highly detailed biological components and circuits, but shared that this work “demonstrates how to systematically explore the biological design space and make biological engineering more predictable. In the future, it will be exciting to generalize this approach to other interactions and components, bringing cells closer to being fully programmable.”
