Week 1 HW: Final Project Ideas

I have always pondered upon the whole medical system that we humans have built over the decades and am always intrigued by the cascade of chemical reactions and pathways that come into action and how that is disrupted when a person is said to have a disease or a disorder.

Standard drug design primarily relies on a puzzle-piece fit to block and interrupt specific medical pathways. I am exploring a hybrid approach that builds upon this established inhibition model by introducing a new complementary mechanism. The question I thought of is : instead of injecting medicine that travels through your whole body, what if we could put a tiny living helper inside you that makes the medicine exactly where it is needed? That helper would be a harmless bacterium—one that already lives in the gut—engineered to produce a therapeutic protein on the spot. Different versions of this helper could be built for different problems: calming inflammation in autoimmune disease, restoring barrier protection in patients whose immune systems are weakened, or even breaking down harmful metabolites in metabolic disorders. This matters because today’s biologic medicines are expensive, require injection, and often suppress the entire immune system.

What do you want to make?

The target product is an orally administered, engineered probiotic bacterium that colonizes a mucosal surface and secretes a therapeutic protein directly at the site of disease, functioning as a “living immune reconstitution platform.” Rather than injecting systemic biologics that suppress the whole immune system, this approach delivers the therapeutic molecule locally—for example, an anti-IL-23 nanobody for inflammatory bowel disease, IL-22 for immunocompromised patients with mucosal barrier injury, or a metabolic enzyme for a condition like phenylketonuria. The motivation is to provide targeted, site-specific immune modulation that avoids systemic toxicity and offers a more accessible alternative to injection-based medicine.

What living system could make it?

The proposed chassis is Escherichia coli Nissle 1917 (EcN), a well-characterized probiotic strain with a long safety record and an established genetic toolkit for therapeutic protein secretion—though maybe other chassis such as Lactobacillus reuteri, Bacteroides thetaiotaomicron, or even a synthetic consortium could prove better suited for specific applications or mucosal sites. The system would integrate three genetic components: a payload expression cassette encoding the therapeutic protein, a secretion system (such as the HlyA Type I secretion system) to export the protein out of the bacterial cell, and optionally an inducible control circuit (e.g., FSY-responsive) for external dosing control. Inputs are dietary nutrients and, if required, a small-molecule inducer; the biological process involves bacterial colonization, transcription and translation of the payload, and active secretion into the local environment; the output is local delivery of a functional therapeutic protein. Maybe future iterations could use cell-free systems for prototyping before committing to a living chassis.

How would automation help you make it?

Automation would primarily be used to screen and optimize genetic circuit variants at the prototyping stage. A liquid-handling robot (such as an Opentrons) would automate the assembly and transformation of hundreds of promoter–RBS–secretion tag combinations, followed by automated liquid handling for protein secretion assays (sampling culture supernatants for ELISA or functional binding assays) and colonization competition assays (co-culturing engineered strains with representative native bacteria). A Cloud Lab could run parallel expression experiments under controlled anaerobic conditions to build a predictive model of payload production, which is essential because the bottleneck is finding the circuit that produces enough protein without burdening the bacterium. Maybe the same automation infrastructure could later be repurposed for personalized strain design, where a patient’s own microbiome data informs which chassis and payload combination would work best for them.

What would it take to make it at a meaningful scale?

For 1,000 patients requiring a daily oral dose, the annual requirement would be on the order of 10¹⁵ CFU, which at high-density fermentation yields translates to only a few hundred liters of culture per year, a scale achievable in a single bioreactor over a handful of batches. After lyophilization, this would be just a few kilograms of powder annually. Per-batch manufacturing time would be approximately one to two weeks (fermentation, harvest, lyophilization, QC), meaning the entire annual requirement could be produced in a matter of weeks. The real bottleneck for clinical translation is not manufacturing volume but genetic stability and biocontainment: ensuring the engineered strain maintains its therapeutic payload over time and cannot spread its genetic material to native bacteria. As a person in the field of bioinformatics, I still have a lot to learn and process on how scaling actually works in the manufacturing level.

What don’t you know yet?

The three biggest unknowns are: (1) long-term genetic stability of the engineered circuit in the competitive mucosal environment, since plasmid loss or mutation would abolish therapeutic production; (2) reliable biocontainment that prevents horizontal gene transfer while still allowing the strain to persist long enough for therapy; and (3) dose predictability, because live bacteria grow, die, and are washed out in ways that administered CFU counts cannot fully control, making it difficult to guarantee consistent therapeutic exposure in patients. Maybe some of these challenges will be solved by advances in genome engineering or inducible kill-switches.