Week 1 HW: Final Project Ideas

Final Project: Light-Activated Post-Harvest Preservation Coating

What do I want to make?

A light-activated, sprayable protein coating that helps small-scale farmers reduce post-harvest fruit loss without needing cold-chain refrigeration. Post-harvest loss is a major bottleneck for smallholder farmers globally, who often lack access to refrigeration and lose large portions of their harvest to spoilage. I’m starting with bananas specifically India is the world’s largest banana producer, and bananas ripen rapidly in tropical conditions with the intent to extend the approach to other climacteric produce (mangoes, tomatoes, etc.) once the core mechanism is validated.

The vision: farmers rehydrate a shelf-stable powder sachet into a liquid protein solution, spray it onto freshly harvested bananas, then expose the crate to a brief flash of blue/UV-A light. The light triggers the coating to cure into a thin, gas-limiting barrier that slows respiration and ethylene-driven ripening a “farmer-controlled optical seal,” rather than relying on an automatic, hard-to-detect spoilage signal like ripening heat (which I originally considered, but it’s too weak and unreliable to distinguish from normal ambient temperature swings). To keep the sachet itself shelf-stable without expensive freeze-drying infrastructure, the formulation would also incorporate tardigrade-derived CAHS (desiccation- tolerance) proteins as a lower-cost way to stabilize the powder for storage and transport.

What living system will I use?

A cell-free protein synthesis (CFPS) platform, likely derived from an industrial E. coli strain, used to manufacture all protein components in a centralized facility rather than spraying any living, engineered organism directly onto food (avoiding a major food-safety/regulatory hurdle). The CFPS system would produce three distinct recombinant proteins:

  1. Silk fibroin-based protein chains — the structural barrier material that forms the actual coating on the banana. Silk fibroin has already been shown to form thin, edible, gas-limiting coatings on fruit.
  2. Photo-reactive tyrosine-containing protein domains, combined with riboflavin (vitamin B2) as a natural, food-safe photoinitiator. When exposed to light, riboflavin generates reactive species that drive crosslinking between tyrosine residues, curing the liquid coating into a solid barrier on command.
  3. Tardigrade-derived CAHS (desiccation-tolerance) proteins — not part of the coating itself, but mixed into the formulation to protect the silk fibroin and crosslinking proteins during drying, so the final product can be shelf-stabilized into a powder sachet without requiring expensive industrial freeze-drying equipment.

All three proteins are recombinantly expressed via CFPS from engineered DNA templates none are extracted from actual silkworms or tardigrades.

How does automation connect (cloud labs, desktop robots, microfluidics)?

The main engineering challenge is tuning the formulation so it cures reliably and quickly under low-energy light, without cracking during rough transit. I’d use a desktop liquid-handling robot (e.g. Opentrons) combined with programmable microfluidics to generate many small-scale formulations varying protein concentration, riboflavin ratio, and pH and test how each responds to different light wavelengths and intensities. Optical sensors could track how quickly each formulation transitions from liquid to solid, letting me map out which combination cures fast enough for real agricultural use. This Opentrons/microfluidics setup is a development and pilot-scale platform. At full agricultural volume, the same optimized formulation would move to larger-volume, parallelized bioreactors with automated purification and fill lines, rather than scaling by running more small-well reactions.

How will I scale it?

For a hypothetical 1,000 kg batch of bananas (~6,667 bananas at ~150g each), assuming roughly 15 mL of a 3% (w/v) protein spray per banana, that’s about 100 liters of spray, requiring roughly 3 kg of pure protein. Cell-free systems can currently yield roughly 1–2 g of protein per liter of lysate, so producing 3 kg would require on the order of 1,500–3,000 liters of lysate achievable with a couple of batch runs in a pilot-scale (~1,000 L) bioreactor over roughly 1–2 days. This 1–2 g/L figure reflects best-case, highly optimized published CFPS systems; typical reported yields are often lower (hundreds of mg/L), so this is an optimistic first-pass assumption rather than a conservative one. These numbers are rough, assumption-based estimates for a first pass, not measured or sourced values.

Real-world precedents

  • Apeel Sciences commercializes a plant-derived lipid coating that extends produce shelf life 2–3x and is FDA GRAS-approved, used commercially on avocados, citrus, and mangos.
  • UBC (2026, Journal of Agricultural and Food Chemistry) developed an agar (seaweed-derived) coating with zinc and tannic acid that kept strawberries fresh 4+ days at room temperature.
  • GreenPod Labs (Chennai, India), founded by Deepak Rajmohan, takes a different approach: instead of a physical coating, their sachets release plant-extract bioactives that activate the produce’s own natural defense mechanisms, extending shelf life 40–80% at 35°C without cold storage — already FSSAI-approved and tested on a range of produce including mangoes, oranges, grapes, tomatoes, and leafy greens. This is a useful contrast to my coating-based approach: it shows that biologically activating existing plant defenses can be as effective as adding an external barrier, and is a real, already-commercialized alternative strategy worth comparing against.

Notably, none of these existing solutions use a deliberate, user-triggered activation step they’re all passive once applied (self-assembly, humidity annealing, or diffusion). The farmer-controlled light trigger in my concept appears to be the most distinct part of this idea relative to what already exists.

What don’t I know yet? (Open questions)

  1. Uneven light exposure: Bananas grow in tightly curved bunches. How would a low-cost light rig reach shadowed areas between bananas in a stacked crate? Uncured spots could rot before the rest of the batch.
  2. Respiration/suffocation risk: I don’t know the safe gas-permeability range for this coating. Too dense a barrier could block oxygen entirely and cause anaerobic fermentation inside the fruit instead of preventing spoilage.
  3. Water quality variability: Since farmers would rehydrate a dry sachet with local water, I’m unsure how variable mineral content or pH in rural water supplies would affect the riboflavin crosslinking reaction and cure time.
  4. Freeze-drying alternative: Whether CAHS-style desiccation proteins could realistically replace freeze-drying for producing a shelf-stable powder at real agricultural scale, not just in lab-scale demonstrations.

References

  • Marelli, B., Brenckle, M. A., Kaplan, D. L., & Omenetto, F. G. (2016). Silk fibroin as edible coating for perishable food preservation. Scientific Reports, 6, 25263. https://doi.org/10.1038/srep25263
  • Omenetto, F., Kaplan, D., Marelli, B., & Brenckle, M. Biopolymer-based preservation of perishable products. US Patent 10,271,561 B2 (Tufts University). https://patents.google.com/patent/US10271561B2
  • Taghizadeh, M., Mohammadifar, M. A., Sadeghi, E., Rouhi, M., Mohammadi, R., Askari, F., Mortazavian, A. M., & Kariminejad, M. Photosensitizer-induced cross-linking: A novel approach for improvement of physicochemical and structural properties of gelatin edible films.
  • Mahmud, J., Muranyi, P., Salmieri, S., Shankar, S., & Lacroix, M. (2024). UV-C-activated riboflavin crosslinked gelatin film with bioactive nanoemulsion for enhanced preservation of fresh beef in modified atmosphere packaging. Foods, 13(21), 3504. https://doi.org/10.3390/foods13213504

Note on AI assistance

I used Claude and OpenAI’s ChatGPT to help brainstorm, refine, and sanity-check this concept, including verifying which proposed mechanisms and citations were scientifically real versus overstated or fabricated. An earlier AI-generated draft included a citation (“Ahmad et al., 2020”) that could not be verified and was likely fabricated — it has been removed and replaced with verified sources above.