Week 1 HW: Final Project Ideas

1. What do you want to make?

I want to biomanufacture targeted Lipid Nanoparticles (LNPs)—microscopic lipid bubbles—carrying temporary messenger RNA (mRNA) and microRNA instructions. These nanoparticles are engineered to specifically target activated scar cells (cardiac fibroblasts) in the heart attack scar bed of adults.

This therapy is designed for post-myocardial infarction (heart attack) patients who are at risk of progressive heart failure. When a heart attack occurs, oxygen starvation kills heart muscle cells (cardiomyocytes). Because adult human hearts cannot naturally regenerate, the body panics and forms a dense, rigid collagen scar to prevent the heart wall from rupturing. However, this stiff scar creates a massive physical and electrical barrier that blocks transplanted stem cells (like iPSC-derived cardiomyocytes) from successfully taking root and surviving.

My project creates a transitional “stall and settle” bridge for cardiac regenerative medicine. Rather than altering the patient’s DNA permanently or surgically cutting out the scar tissue, the biomanufactured nanoparticles deliver temporary mRNA/microRNA “instruction notes.” These instructions briefly turn down collagen production in the scar cells, softening the existing hard matrix right as iPSC-derived cardiomyocytes are introduced. This mimics natural regenerative species (like zebrafish and neonatal mice), keeping the structural integrity of the heart wall intact while stalling time for the new stem cells to safely integrate, receive blood supply, and rebuild functional heart muscle.

My motivation combines systemic health challenges, utilitarian biomedical engineering, and a little bit of personal experience. Globally, cardiovascular disease remains the number one cause of death in many countries. The heart is the primary engine of the body, responsible for pumping oxygen-rich blood to every other organ. When the heart fails, it triggers a domino effect where other organs slowly starve and fail as well. As the global population ages with passing generations, finding effective solutions for post-heart attack remodeling is becoming increasingly urgent. Beyond the statistics, watching a loved one or patient slowly lose their organ function and independence to heart failure is heartbreaking. It robs people of their vitality, mobility, and dignity long before life ends. While volunteering, I met a patient who suffered from recurrent heart attacks while serving as the primary caregiver for his terminally ill father. He expressed intense fear that his weakening heart would fail before he could finish caring for his parent. That conversation showed me that heart disease strips away a person’s capability to care for those they love too. It’s not just elders but also family members who may be the primary provider. Also, designing a biomanufactured solution that gives damaged hearts a realistic, safe path to heal is a problem I find intriguing as it opens more possibilities for other organ regeneration.

2. What living system could make it?

To be completely honest, I don’t have a background in science, so setting up an experiment like this was definitely confusing, and I’m still reading through research to figure it all out. So, I might have a few many things that are wrong. My project builds directly on the stem cell research pioneered by Dr. Chuck Murry. As Dr. Murry demonstrated in his work on regenerating heart muscle, we can take human pluripotent stem cells (iPSCs), differentiate them with high efficiency into cardiac muscle cells (iPSC-cardiomyocytes), and transplant them into injured heart tissue to re-seed the heart wall. However, one of the biggest clinical hurdles stem cell therapies face after a heart attack is the stiff, dense scar tissue left behind. If the scar bed is too rigid, transplanted stem cells struggle to integrate, take root, and achieve the synchronous beating and electromechanical coupling needed to restore function without causing arrhythmias. To solve this, my system focuses on temporarily softening that scar tissue right when the stem cells are introduced. For my biomanufacturing living system, I chose a cell-free enzymatic system (In Vitro Transcription) paired with a microfluidic lipid assembly platform, rather than growing whole living cells in big tanks. By isolating just the essential purified enzymes, we can produce our targeted matrix-softening mRNA and microRNA payloads directly—without having to feed living cell cultures, clean up cellular waste, or worry about viral contamination.

This part I got alot of help from the papers and Gemini The setup works by feeding in (1)Linearized DNA templates encoding matrix-softening signals (like matrix metalloproteinases [MMPs] to gently break down rigid collagen, and microRNA-29 to temporarily suppress new collagen synthesis) (2)Modified NTP building blocks and T7 RNA polymerase to synthesize the RNA strands. (3)Targeting lipids conjugated with peptides that specifically recognize activated cardiac scar cells (myofibroblasts). Inside the cell-free reactor, the enzymes build the RNA, and a microfluidic mixer immediately wraps it into uniform lipid nanoparticles (LNPs). The output is a batch of sterile, targeted LNPs designed to give temporary scar-softening instructions to cardiac myofibroblasts at the transplant site. My main assumption and biological mechanism— cell-free principles in Karim & Jewett (2021) and research on cardiac scar reprogramming from Dr. Conrad Hodgkinson’s lab at Duke and the University of Sydney (2026)—is that synthetic mRNA naturally degrades within 24 to 72 hours. That short lifespan acts as a built-in safety timer: it pauses collagen production and softens the matrix just long enough for Dr. Murry’s transplanted iPSC-cardiomyocytes to settle in and couple with host tissue, before normal scar repair naturally resumes to prevent wall rupture.

3. How could automation help you make it?

Automation seems to cut through a lot of tedious manual work across two key parts of the project: Before making a working nanoparticle at scale, hundreds of different combinations of RNA sequences (MMPs, miR-29) and lipid formulations have to be tested to find the exact right ratio. I would use a desktop liquid-handling robot (like an Opentrons OT-2) to automate mixing reactions in 96-well plates. The robot gives us sub-microliter precision, speeds up testing, and makes sure our cell-free transcription assays are completely reproducible. Once we have the right recipe, we need to package the RNA into tiny lipid bubbles (LNPs). Pushing fluids through microfluidic channels manually with a syringe is risky because pushing too hard or unevenly creates turbulent flow or damages the chip. Instead, I would use an automated microfluidic platform (like a Precision Nanosystems or KNAUER LNP Skid). This system automatically controls the exact flow rates, pressures, and mixing speeds of the liquid RNA and organic lipid streams as they converge in microscopic channels. Because nanoparticle self-assembly is super sensitive to fluid dynamics, automating the flow rates guarantees uniform particle sizes (~80–100 nm) and high encapsulation efficiency without batch-to-batch variation.

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

What I understood after talking with AI and going through readings on this topic, biomanufacturing scale is all about working backward from the patient dose while accounting for real-world biological losses. To make a meaningful impact, I defined my target scale as treating 10,000 post-myocardial infarction patients per year who are receiving stem cell therapy. A standard therapeutic dose for targeted lipid nanoparticles (LNPs) in cardiac applications is roughly 1X10^{13 active particles per patient. However, because most particles in a real human body get cleared by the liver or break down before reaching the heart, I assumed a 20% target delivery efficiency (meaning 80% is lost). That means we actually need to manufacture 5X10}13 LNPs per patient, bringing our total annual requirement for 10,000 patients to 5X10^17 total LNPs per year. Dividing our total required particles by this yield shows that we need to produce 500 Liters of bulk LNP liquid formulation per year. When it comes to manufacturing time, what I learned from the readings is that the automation tools used during early prototyping are completely different from what you use for full-scale production. During the R&D phase in the lab, when we are testing small 10 mL to 100mL batches using a desktop liquid handler (like an Opentrons OT-2) paired with a benchtop microfluidic mixer, setting up a cell-free in vitro transcription reaction and encapsulating it into LNPs takes under 2 hours per trial batch. But once we move to full clinical production for our 500-liter annual supply, we switch to a continuous-flow industrial microfluidic skid operating at a throughput of roughly 500mL/min (30Liters/hour). Running 500 Liters through that system takes approximately 16.7 hours of continuous fluidic operation. Even after factoring in extra time for tangential flow filtration purification and sterile vial packaging, the entire year’s supply for 10,000 patients could be manufactured in under 3 to 5 operational days.

5. What don’t you know yet?

The biggest unknown is figuring out the exact types and ratios of mRNA and microRNA constructs needed to temporarily soften the dense collagen scar without interrupting or damaging the incoming iPSC-cardiomyocyte stem cells. If we break down the scar tissue too aggressively, we risk causing the left ventricular wall to rupture, but if we don’t break it down enough, Dr. Chuck Murry’s transplanted stem cells won’t be able to integrate or form the electromechanical gap junctions (Connexin-43) required to beat synchronously with the host heart. Additionally, while I’m assuming our peptide-guided lipid nanoparticles will target activated cardiac myofibroblasts, I don’t know if the liver will filter most of them out before they even reach the heart. To actually validate this idea in the lab, I would need to measure the mechanical stiffness of human cardiac tissue after treatment, run biodistribution assays to track LNP clearance, and perform electrophysiological mapping on co-cultured scar and stem-cell tissues to ensure the cells couple properly without causing dangerous arrhythmias.

Here are the reading/videos:

1. Can we regenerate heart muscle with stem cells? | Chuck Murry
2. Heart health enters the future with a $28.7 million gift from the Marcus Foundation
3. What Happens During a Heart Attack | Anatomy of a Heart Attack
4. Cardiac fibroblasts in heart failure and regeneration - PMC
5. Duke Team Turns Scar Tissue into Heart Muscle Without Using Stem Cells | Duke Health
6. Reprogramming Scar Tissue to Heal the Heart: Dr. Conrad Hodgkinson Receives NHLBI R01 Award | Duke Department of Medicine
7. Scientists discover the human heart can regrow muscle after a heart attack | ScienceDaily
8. Human heart regrows muscle cells after heart attack, world-first study shows - The University of Sydney

Note: I used Gemini to brainstorm my ideas before answering the question. Specifically, the AI helped translate my strategy—creating a transitional “stall and settle” stage using temporary mRNA/microRNA signals to soften post-infarction heart scars while iPSC-derived cardiomyocytes integrate—into technical engineering terms. Gemini assisted in evaluating biomanufacturing platforms by comparing living cell cultures against cell-free enzymatic synthesis (In Vitro Transcription), clarifying the biological composition of cardiac scars (such as Collagen type I and myofibroblast activity), identifying suitable automation platforms (Opentrons liquid handlers for R&D screening and microfluidic mixers for lipid nanoparticle assembly), and structuring back-of-the-envelope scale calculations and bottleneck analyses (such as off-target liver accumulation and cell-free enzyme costs). Additionally, I used the AI to help organize my raw ideas into logical responses for Prompts 1 through 5. I remain completely responsible for all final concepts, calculations, and written content presented in this assignment.

Week 2 HW: DNA Design

Designing an MMP-13 mRNA Template for Cardiac Scar Remodeling

Project overview: from Homework 1 to Homework 2

In Homework 1, I proposed investigating whether the collagen-rich scar that forms after a myocardial infarction (heart attack) could be remodeled in a controlled way to make the damaged heart more receptive to future regenerative treatments, including approaches using stem-cell-derived cardiomyocytes. This idea is related to the broader field of cardiac regeneration, including work associated with Dr. Charles (Chuck) Murry.

Scar tissue is not simply a problem to remove: early after an infarction, it helps hold the injured heart wall together. My question is whether carefully timed and localized remodeling of an established scar could be useful without compromising its mechanical integrity. The literature on myocardial fibrosis and matrix metalloproteinases makes this a biologically plausible question, but also highlights important risks (MMPs in MI and heart failure; myocardial fibrosis pathways).

My proposed manufacturing route is cell-free in vitro transcription (IVT) to make mRNA, followed conceptually by lipid nanoparticle (LNP) delivery to mammalian cells. In Homework 2, I focused on designing the DNA template that could be used to make that mRNA. I am still learning the underlying molecular biology, so this is a first design.

Concept: DNA template → cell-free IVT → MMP13 mRNA → proposed LNP delivery → mammalian cell → MMP-13 protein → possible extracellular-matrix remodeling.

Part 0: Setup in Kernel

I created a Kernel repository for this project: Nam-mmp13-cardiac-scar. I used Kernel to collect genetic parts, assemble a construct, generate a codon-optimized coding sequence, and run a sequence-level biosecurity screen. I also used NCBI BLAST for sequence identification.

Part 1: Expression target

Molecule. I selected human matrix metalloproteinase-13 (MMP-13), also called collagenase-3. It is an enzyme involved in extracellular matrix (ECM) turnover. I considered other MMPs, including MMP-14, but chose secreted MMP-13 as a working model for extracellular collagen remodeling.

Sequence and source. The target is the human MMP13 gene (NCBI Gene ID 4322), with reference mRNA NM_002427.4, reference protein NP_002418.1, and UniProt protein P45452 (MMP13_HUMAN). The reference protein is 471 amino acids long. My Kernel coding sequence was generated from the reference protein sequence rather than copied unchanged from the natural human nucleotide sequence (NCBI MMP13; RefSeq NM_002427.4; [UniProt P45452] (https://www.uniprot.org/uniprotkb/P45452/entry)).

Host/chassis. The immediate manufacturing system is a cell-free T7 IVT reaction: it transcribes a DNA template into mRNA. The later translation host would be a mammalian target cell, which would use the delivered mRNA to make MMP-13 protein. These are two different stages, which is why I used a human translation context for codon optimization.

Destination. The mRNA would need to reach the inside of mammalian cells. The resulting MMP-13 protein would need to be processed and secreted to reach the extracellular collagen matrix. The native MMP-13 precursor contains N-terminal secretion-related information. Whether the protein would be secreted and activated appropriately remains to be tested.

Part 2: Genetic parts

Important design distinction: A T7 promoter is for making the RNA in vitro, while the mammalian UTRs and CDS are intended to support what happens after that RNA enters a mammalian cell. The assembled parts do not by themselves establish correct RNA end processing, capping, translation efficiency, or clinical suitability.

Part 3: Assemble the construct

In Kernel, I assembled a construct named MMP13-IVT-Construct in the following 5′→3′ order:

T7 promoter → HBA1 5′ UTR → MMP13-Human-Optimized-CDS → HBB 3′ UTR → poly(A) tail

Kernel displayed a total length of 1,638 bp. The individual lengths were 23 bp, 37 bp, 1,416 bp, 132 bp, and 30 bp, respectively. I inspected the part order and orientation in the construct map. The optimized CDS starts with ATG, and the optimizer settings specified a TAA stop codon. Kernel showed a translated coding region; a complete independent frame and junction verification would still be prudent before synthesis.

In particular, the displayed construct has no selected backbone, and the suitability of the 5′/3′ ends and poly(A) design still needs to be evaluated.

Figure 1. MMP13-IVT construct assembled in Kernel.

Part 4: Optimize and quality-check

4.1 Codon optimization

I used Kernel’s Coding Sequence Optimization tool with the human MMP-13 amino-acid sequence from UniProt P45452 as input. I selected Homo sapiens as the expression host and Kernel’s SIMPLE_0_3 optimizer. The tool generated a 1,416-bp DNA CDS with a TAA stop codon. I did not specify restriction-site avoidance or CpG avoidance.

Codon optimization selects DNA codons for an intended translation context while aiming to preserve the protein’s amino-acid sequence. I chose a human host because mammalian cells would translate the delivered mRNA; the cell-free T7 IVT stage only makes the RNA.

Figure 2. MMP13 codon optimization results in Kernel.

4.2 Sequence search / BLAST

I submitted the 1,416-bp optimized CDS to NCBI BLASTN against the core nucleotide database (core_nt) (NCBI BLAST). The results included human MMP13 / collagenase-3 sequences, including the reference mRNA NM_002427.4. The human reference match shown in my results had 100% query coverage, an E-value of 0.0, and approximately 75.85% nucleotide identity.

Figure 3. NCBI BLASTN results for the codon-optimized MMP13 sequence.

4.3 Biosecurity sequence screening

I used Kernel’s Biosecurity Sequence Scanner, powered by SecureDNA, on the optimized MMP13 CDS. It reported: “No flagged sequences detected.” This means the submitted sequence did not trigger that screening tool’s biosecurity flags. Also, this result was obtained for the CDS, not necessarily a separate scan of the entire 1,638-bp construct.

Figure 4. Kernel SecureDNA biosecurity screening results.

Part 5: From design to build

How I would build it. For an initial prototype, I would propose ordering the approximately 1.64-kb designed cassette as commercially synthesized DNA, subject to a provider’s sequence review. Because the cassette is relatively short and has only a few parts, this seems simpler than generating it from a large pool of chip-synthesized oligonucleotides. If synthesizing the complete cassette proved difficult, I would consider separate verified fragments and an assembly strategy such as the one demonstrated by Gibson et al. (2009). Kosuri et al. (2010) is more relevant to parallel synthesis of many designs, which could become useful if I later compared many UTR or coding-sequence variants.

Synthesis constraints. A digital sequence is not guaranteed to be reproduced perfectly as physical DNA. Carr et al. (2004) describe error correction in de novo DNA synthesis, emphasizing why sequence fidelity matters. I have not yet completed a provider-level assessment of GC content, repeats, or difficult sequence features, so I cannot claim that my cassette is trouble-free to synthesize. I would require sequence verification of the delivered DNA and reconsider a fragment-based design if the provider flagged difficult regions.

Prototype versus scale. For proof of concept, the priority would be a small amount of verified DNA template and evidence that it yields the intended RNA. At scale, my proposed product would be mRNA, not the DNA itself: the main manufacturing challenge would shift toward reproducible template quality and IVT-derived RNA quality. This distinction follows my Homework 1 concept, but neither the DNA production process nor the mRNA manufacturing process has been experimentally validated here.

Part 6: What I do not know yet

1. Will the mRNA yield properly processed and secreted MMP-13? Kernel can show a plausible CDS but cannot establish successful expression, folding, secretion, activation, or activity. A future mammalian-cell study would compare mRNA abundance, MMP-13 protein production, secretion into the extracellular medium, and relevant matrix-degrading activity. A lack of secretion or activity would mean the expression design needs revision.

2. Could MMP-13 weaken the heart scar? The scar is structurally important, especially during early healing. Too much or poorly timed collagen breakdown might increase damage rather than improve regeneration. Future work would need to evaluate ECM remodeling alongside tissue integrity and cardiac function, rather than treating greater collagen degradation as automatically beneficial. This is the central safety uncertainty in the project (MMPs in MI and heart failure).

3. Would mRNA delivery reach the intended cardiac cells? A successful DNA and RNA sequence would not guarantee targeted delivery. Future research would compare distribution and cellular uptake of candidate delivery systems, including off-target exposure. I would need to establish where the RNA and protein go before making any claim of therapeutic efficacy.

AI use disclosure

I used ChatGPT (OpenAI) as a learning and writing assistant to explain basic molecular-biology concepts, help translate my Homework 1 proposal into a Kernel design, organize my observed Kernel and BLAST results, and structure this write-up. I reviewed the screenshots and supplied the experimental-tool outputs reported above. AI assistance does not substitute for sequence verification or experimental validation.

References and background resources

Genetic design and DNA synthesis (assigned readings)

• Carr, P. A., et al. (2004). “Protein-mediated error correction for de novo DNA synthesis.” Nucleic Acids Research, 32, e162. https://doi.org/10.1093/nar/gnh160
• Gibson, D. G., et al. (2009). “Enzymatic assembly of DNA molecules up to several hundred kilobases.” Nature Methods, 6, 343–345. https://doi.org/10.1038/nmeth.1318
• Kosuri, S., et al. (2010). “Scalable gene synthesis by selective amplification of DNA pools from high-fidelity microchips.” Nature Biotechnology, 28, 1295–1299. https://doi.org/10.1038/nbt.1716

Sequence and software resources

• NCBI Gene: Human MMP13 (Gene ID 4322)
• NCBI RefSeq: NM_002427.4
• UniProt: MMP13_HUMAN (P45452)
• NCBI BLAST
• SecureDNA

Heart attack, fibrosis, and regenerative-medicine background

The following are background links I collected for this project. Not all were used to support a specific claim above, and their relevance should be checked before using them as formal scholarly citations.

• Khan Academy — Healing after a heart attack (myocardial infarction)
• Cellular mechanisms of tissue fibrosis: contributory pathways leading to myocardial fibrosis
• Matrix Metalloproteinases in Myocardial Infarction and Heart Failure
• PMC article: PMC11090942
• PMC article: PMC3050537
• PMC article: PMC13026035
• PMC article: PMC12439134
• Stanford News — Cardiac fibrosis research (April 2025)
• PMC article: PMC8063206
• PMC article: PMC8355135

Final Project