Week 1 HW: Final Project Ideas

1 What you would like to biomanufacture and why?

Describe the target product or output you would like to make. Why are you interested in making it? Your motivation might be utilitarian—for example, addressing a problem in health, sustainability, food, energy, or manufacturing—but it does not need to be. Your project might instead have a social, cultural, artistic, or personal motivation. Clearly articulate what you want to make, who or what it is for, and why you think it is interesting or important.

Target: a transient mRNA-based partial cellular reprogramming platform expressing OCT4, SOX2 and KLF4 (OSK) in aged human cells.
The target product is a research-grade messenger RNAs encoding the transcription factors OCT4, SOX2 and KLF4. The immediate use case is controlled laboratory research in aged human cell and tissue models; a longer-term goal would be to understand whether transient delivery could support regenerative applications.
The biological motivation is to investigate whether aging-associated cellular states can be shifted toward a younger state without fully reprogramming cells into pluripotent stem cells. Full reprogramming can reset many age-associated molecular features, but it also erases differentiated cell identity. Partial or transient reprogramming instead attempts to capture early rejuvenation-associated changes while stopping before full dedifferentiation.
The potential clinical benefits would be substantial if the underlying premise holds. Aging is the dominant risk factor for most chronic disease, and the geriatric-science hypothesis proposes that intervening in aging biology could therefore delay several age-related conditions at once. That hypothesis remains to be demonstrated in humans and is treated here as motivation rather than as established fact. Among the possible targets, physical decline is the most tractable to measure, while cognitive decline is arguably the most consequential.
mRNA is attractive because expression is transient and non-integrating. The RNA is translated in the cytoplasm and subsequently degraded, so the intervention does not require permanent insertion of the OSK genes into the genome. This feature is especially relevant to partial reprogramming, where duration and magnitude of factor expression are expected to determine whether a cell remains within a useful rejuvenation window or progresses too far toward dedifferentiation.
Sakar et al. has shown that transient, non-integrating mRNA expression of reprogramming factors can improve several aging-associated molecular and functional readouts in human cells while preserving cell identity under the tested conditions ¹. Two qualifications matter for this proposal. First, that study used a six-factor cocktail (OCT4, SOX2, KLF4, c-MYC, LIN28 and NANOG, abbreviated OSKMLN), not OSK alone; restricting the cocktail to three factors is a deliberate simplification of this project and is not itself supported by that the referenced publication. Second, Gill et al. demonstrates substantial epigenetic rejuvenation before complete reprogramming, but does so using a doxycycline-inducible, lentivirally integrated polycistronic OKMS construct rather than transient mRNA, so it supports the biological premise and not the delivery format proposed here ². Cipriano et al. reviews the broader evidence that rejuvenation can precede full reprogramming ³.

2 What living system would you use?

Propose at least one living-system manufacturing chassis that could produce your target—for example, a microbial, mammalian, or cell-free system. At a high level, describe how the system would make your target product. What goes into the system? What biological process or processes occur? What comes out? You do not need to design the complete biological system yet. Make reasonable assumptions where necessary and identify them.

Proposed chassis: E. coli for DNA-template manufacture, coupled to a cell-free in-vitro transcription (IVT) system for the final mRNA product.
I would use Escherichia coli as the main living manufacturing chassis. The bacterium would not be asked to produce the final therapeutic-style mRNA directly. Instead, it would serve as an efficient biological factory for amplifying plasmid DNA templates encoding OCT4, SOX2 or KLF4.
Why not a mammalian chassis. Human cells are the most biologically analogous living system and would produce the mRNA in its native cellular context, so the alternative deserves stating explicitly rather than being passed over. It is rejected here on three grounds. First, product definition: a human cell produces tens of thousands of distinct RNA species, and isolating one defined, full-length, capped and tailed transcript from that background is far harder than transcribing it from a single linearized template in a clean enzymatic reaction. Second, cost: mammalian culture carries substantially higher operating expense per litre (serum-free media, growth factors, CO2-controlled incubation) and higher capital expense (single-use bioreactor trains, dedicated clean-room classification) than bacterial fermentation. Third, robustness: doubling times are roughly an order of magnitude longer and contamination consequences more severe, which lengthens every design-build-test cycle. Because E. coli is asked only to copy DNA and never to make the product, nothing about the final mRNA requires a mammalian chassis.
The principal inputs are engineered DNA constructs and bacterial growth media. Inside the cells, normal bacterial growth and DNA replication amplify the plasmid templates. The resulting DNA is then isolated and transferred to a cell-free in-vitro transcription system. In IVT, an RNA polymerase reads the DNA template and synthesizes the corresponding mRNA. The RNA is subsequently purified and prepared for research use.
This hybrid architecture separates two jobs: a living organism efficiently copies the DNA template, while a cell-free enzymatic system produces a defined RNA product. That is more controllable than trying to extract one specific mammalian mRNA from the complex mixture of RNAs naturally produced inside a cell.

Assumption: The project treats the living chassis as one stage of a hybrid biomanufacturing process. The final mRNA is produced cell-free because that is the standard and more controllable production logic for defined synthetic mRNA.

3 How would you use automation?

Identify at least one automation technology that you would combine with your living system. This could include a large-scale Cloud Lab, a desktop liquid-handling robot such as an Opentrons, a programmable microfluidics platform such as Nuclera’s, or another automation technology. Describe what part of the biomanufacturing workflow you would automate and why. For example, you might automate liquid handling, culturing, assembly, screening, sampling, measurement, optimization, or another step.

Automation platform: an Opentrons Flex liquid-handling robot, ideally equipped for high-throughput 96-well workflows.
The first automation target would be DNA preparation and small-scale construct screening. Different mRNA designs may need to be compared, for example alternative untranslated regions or alternative ratios of the three factors. A liquid handler could prepare many small reactions in parallel and reduce variability caused by manual pipetting.
The second target would be cell-free mRNA production. The robot could standardize repetitive liquid-transfer steps during the setup of multiple IVT conditions and prepare samples for downstream purification or analysis. Opentrons offers a 96-channel Flex pipette designed to address an entire 96-well plate, making it suitable for high-throughput plate workflows ⁴.
The third target would be sampling and QC preparation. The robot could route aliquots from each manufacturing condition to measurements of RNA concentration, integrity and expression performance, with the results linked to the corresponding production parameters. This creates a reproducible design-build-test-learn loop rather than a series of disconnected manual experiments.

Baseline:
Automation as reliable pipetting. The three targets above describe the conventional justification for a liquid handler - it pipettes faster, more reproducibly and for longer than a person, and it removes operator-to-operator variability. That alone is worth having, but it is the weakest use of the instrument. Opentrons is most useful for development, optimization and parallel QC; at manufacturing scale the workflow would move toward larger controlled IVT vessels and automated downstream purification rather than multiplying hundreds of microwell reactions.

Recommendation: automation as a combinatorial screening engine.
The stronger argument is that a liquid handler makes it cheap to run many slightly different conditions concurrently, which converts a linear optimization campaign into a parallel one. Two linked plate layouts are proposed.

Plate layout A - process-side (IVT) screen.
One 96-well plate encodes a factorial combination of construct and reaction variables: 5’ and 3’ UTR variant, ARCA versus enzymatic capping, modified-nucleotide content, and magnesium and NTP concentration. Each well is an independent IVT reaction; the readouts are yield, RNA integrity, capping efficiency and double-stranded RNA content. The 96-channel Flex pipette addresses the whole plate in a single transfer step ⁴, so a 96-condition screen costs roughly the hands-on time of a single reaction.

Plate layout B - biology-side dose and duration screen.
A second 96-well plate crosses the variables that define the reprogramming window: number of consecutive daily transfections (2, 4 or 6), mRNA dose (0.25x, 0.5x, 1x, 2x of the standard per-well amount), OSK molar ratio (four variants) and donor cell line (three aged plus one young reference), in duplicate. The bracketing is taken from Sakar et al., where two daily transfections already produced measurable effects, four was the standard regimen, and twelve to fifteen were required to generate iPSC colonies; conditions of two to six therefore sit inside the window below the point of no return. The full crossing is larger than one plate, so it is run as a staged fractional factorial rather than exhaustively ¹.

Linking the two.
Every well in both layouts carries a barcode that ties its production parameters to its RNA QC result and then to its cell-model readout, so process variables and biological outcomes are analysed in one dataset rather than two. This is what turns the design-build-test-learn loop from a diagram into an experiment: each plate cycle tests roughly a hundred hypotheses instead of one, and the next design set is chosen from the response surface rather than from intuition.

Second iteration: would microfluidics make sense?
For the synthesis step itself, largely no. Droplet-based or continuous-flow IVT remains a research technique, RNA recovery from droplets is poor, and the physical volumes involved here are small enough (see section 4) that miniaturizing transcription solves a problem the project does not have. For the steps on either side of it, yes, and in two places. First, acoustic or microfluidic dispensing at nanolitre to low-microlitre scale would cut reagent cost per condition in the screens above by roughly an order of magnitude, which matters when the point is to test hundreds of variants rather than to make material. Second, and more importantly, microfluidic mixing is the standard method for lipid-nanoparticle formulation, and delivery is named in section 5 as the leading technical unknown. A microfluidic formulation module is therefore the recommended second-iteration addition - not to make more RNA, but to make the RNA deliverable and to screen LNP compositions with the same parallelism applied to the mRNA itself.

4 How much would you need to make, and how long would it take?

Perform a back-of-the-envelope calculation estimating what it would take to manufacture your target at a meaningful scale.

4a How much?

Define a scale relevant to your project and calculate approximately how much of your target product you would need to manufacture. For example, if you are producing a medicine, you might estimate the amount required to treat 1,000 people, everyone in Massachusetts, or everyone in the world. If you are producing a building material, you might calculate the amount required for one structure or an entire neighborhood. If you are producing an art or design artifact, define an appropriate scale yourself. Clearly show your calculation, including units, assumptions, and sources for important numbers.

Defined meaningful scale: enough purified OSK mRNA for 1,000 standardized research experiments. There is no established human dose for OSK-mRNA partial reprogramming, so a patient-scale calculation would require many unsupported assumptions. I therefore use a research-scale benchmark and define the unit of work physically so that the figure can be checked rather than taken on trust.

Definition of one standardized experiment. One experiment is one treatment condition applied to one well of a 6-well plate, receiving four consecutive daily transfections. The per-transfection amount is not invented: Sakar et al. states that cells were transfected using Lipofectamine MessengerMAX according to the manufacturer’s protocol, and that protocol specifies 2,500 ng of mRNA per well in the 6-well format [1, 7]. Sakar et al. further specifies the regimen as daily transfection for four consecutive days ¹. Four transfections at 2.5 micrograms therefore give 10.0 micrograms of total OSK mRNA per experiment, and this value is derived from the materials and methods of the cited work rather than assumed.

Two experiment formats.
The 10-microgram unit above is the confirmation format, sized so that each condition yields enough cells for multi-omic readouts such as methylation-clock analysis, transcriptomics and mitochondrial function. Screening does not need that many cells. In the 96-well format the same protocol specifies 100 ng per well ⁵, so a four-day screening condition consumes 400 ng - twenty-five times less material per condition. The campaign is planned as a mixture of both: approximately 10,000 screening wells at 400 ng (4 mg) plus approximately 600 confirmation wells at 10 micrograms (6 mg), which preserves the 10 mg total while making explicit that most of the conditions, and therefore most of the information, come from the cheaper format.

Recovery assumption of about 70%.
The 70% figure is anchored in reported chromatographic performance. Krusic reports recovery of purified mRNA above 85% for ion-pair reverse-phase step elution on an 8 mL column, above 76% in linear-gradient mode, and 69% for the same process translated to an 80 mL column ⁶. The 70% used here corresponds to that scale-translated figure. It should be read as the recovery of a single polishing step, not of the entire downstream train; if capture, DNase treatment, concentration and buffer exchange are included, an end-to-end recovery closer to 40% is realistic. A sensitivity case is given below.

DNA template requirement.
The bacterial stage is sized by the same calculation. The cited kit uses approximately 1 microgram of linearized template per standard 20-microlitre reaction ⁷, so 715 reactions require roughly 0.7 mg of linearized plasmid. That is within the output of a single maxiprep from a few hundred millilitres of E. coli culture, per construct. This confirms that the living chassis is correctly sized for the campaign and is not the limiting stage.

Approximately 715 reactions correspond to about eight 96-well plates, or roughly 2.5 plates for each of the three factors, since the RNAs are produced in separate plate groups and combined only after independent purification and quality assessment. Sensitivity: at a pessimistic 40% end-to-end recovery the crude requirement rises to 25 mg, about 1,250 reactions or thirteen plates, and the template requirement to roughly 1.25 mg of linearized plasmid per construct. The conclusion of this section is unchanged under either case.

Process steps assumed but not yet specified.
Three steps are required for the product to be usable and are not captured in the reaction count above. First, the plasmid template must be linearized downstream of the coding sequence for run-off transcription. Second, the mRNA requires a poly(A) tail; the kit cited in ⁷ is the ARCA kit without tailing, so either the tailing variant of that kit or a separate poly(A) polymerase step must be added, and untailed mRNA should not be assumed functional in primary cells. Third, modified nucleotides, or an interferon-pathway inhibitor such as B18R, are needed to make repeated daily transfection of primary human cells tolerable; this changes nucleotide cost and typically reduces yield relative to the unmodified maximum.

Quality attributes.
Double-stranded RNA, truncated transcripts and residual DNA template are the impurities that most directly reduce potency, because they trigger innate immune sensing and consequent translational shutdown ⁶. A dsRNA-removal polishing step is therefore treated as part of the process rather than optional. Because the template is produced in E. coli, residual host DNA and bacterial endotoxin also require specification and release testing before the material enters primary-cell experiments.

4b How long?

Using your proposed living system and automation technology, estimate how long it would take to manufacture that amount. Consider quantities such as production rate, yield, volume, number of parallel experiments or reactors, automation throughput, or whatever parameters are most appropriate for your system. You may discover that the automation technology you would use to prototype your system is very different from what you would use to manufacture at scale. If so, describe both. Again, we are looking for a reasonable engineering estimate. Show your assumptions and calculations.

Two automation systems, not one.
The prototype campaign and a hypothetical manufacturing campaign use different equipment, and they are limited by different things. The prototype path is built around plate-based liquid handling and is limited by the number of plate-handling cycles; the manufacturing path is built around a single stirred reaction and a chromatography skid, and is limited by quality control rather than by production. Both are costed below for the same 14.3 mg crude target from section 4(a).

Path A - prototype campaign on the Opentrons Flex

Throughput assumptions. Bacterial growth: E. coli in rich medium doubles in roughly 20-30 minutes, so a 1:100-inoculated 150 mL culture reaches harvest density overnight (about 16 hours); one maxiprep per construct yields 300-750 micrograms against the 240 micrograms required ⁸. Liquid handling: the 96-channel Flex head addresses all 96 wells in a single transfer ⁴, so one IVT plate needs roughly four plate-wide operations plus a deck exchange, estimated at 15 minutes per plate. Transcription: approximately 2 hours at 37 C ⁷, with all eight plates incubating concurrently off-deck, so incubation is paid once rather than per plate. Purification: magnetic-bead clean-up processes one plate per magnetic module cycle at roughly 40 minutes per plate, which is the only step that does not parallelise across plates.

Bottleneck.
The rate-limiting step is not transcription but plate-wise bead purification: 5.3 hours against 2.25 hours of transcription. Adding a second magnetic module, or switching to a single pooled column clean-up once the plate identities are no longer needed, roughly halves the post-transcription time and is the single highest-value change to the prototype workcell. Note also that 20-microlitre reactions in an unsealed 96-well plate held at 37 C lose volume to evaporation, so a temperature-controlled on-deck module with plate sealing is assumed rather than open-plate incubation.

Path B - manufacturing-scale campaign

Process assumptions.
At manufacturing scale the plate format disappears. The same 14.3 mg crude target is made in a single stirred, temperature-controlled reaction rather than in 715 microwell reactions. Kit chemistry corresponds to roughly 1 g/L (20 micrograms in a 20-microlitre reaction ⁷), giving 14.3 mL of reaction volume; optimised process chemistry reaches 5-15 g/L ^{6 9}, giving 2.9 mL at the low end. Either fits in a single small vessel. Purification moves from magnetic beads to monolithic ion-pair reverse-phase chromatography, for which Krusic et al. provides directly usable sizing ⁶: a load challenge of 0.6 mg RNA per mL of column corresponds to 86% of dynamic binding capacity, implying a capacity near 0.7 mg/mL, with step-elution recovery of 69-85% depending on column scale.

Automation at scale is different in kind.
Path B uses almost no liquid handling. Its automation is process control: a single-use fermenter with automated feed and dissolved-oxygen control for the template stage, a chromatography skid with inline UV and conductivity monitoring and automated fraction gating for purification, and automated in-process sampling into the QC train. The Opentrons has no role in Path B except in preparing QC assay plates.

Conclusion of the timing analysis.
For this campaign the manufacturing path is the faster of the two, because 14.3 mg is a single small reaction. The prototype path is therefore not chosen for speed or for mass throughput; it is chosen because the scientific question requires many different RNAs and many different treatment conditions rather than a large quantity of one. Prototype automation is justified by variants per run, manufacturing automation by grams per day, and conflating the two is the most common way this kind of estimate goes wrong.

5 What don’t you know yet?

After completing your analysis, identify the 2–3 biggest technical unknowns, bottlenecks, or assumptions in your proposed biomanufacturing system. What would you need to learn, measure, or experimentally determine to know whether your idea could actually work?

The critical uncertainties are biological and delivery-related, not simply whether enough RNA can be produced.

Delivery efficiency and tissue specificity
The mRNA must enter the intended cells, escape endosomes, reach the cytoplasm and produce adequate OSK protein expression. What must be measured: fraction of target cells reached, expression magnitude, expression duration, repeat-dose behavior and off-target delivery.

The safe and useful OSK expression window
The most important unknown is whether a reproducible dose-duration region exists in which aging-associated features improve without unacceptable dedifferentiation, loss of cell identity or abnormal proliferation. What must be measured: time-resolved factor expression, cell-identity markers, pluripotency-associated programs, growth behavior and genomic stability.

What counts as genuine rejuvenation
A change in a single molecular marker would not establish functional rejuvenation. What must be measured: multiple independent aging-associated readouts, including epigenetic/transcriptomic state, mitochondrial function, DNA-damage or senescence-associated features, plus cell-type-specific function and adverse phenotypes.

Cocktail composition and product quality
The closest mRNA-based precedent ¹ used six factors (OSKMLN); restricting to OSK is an untested simplification of this project, not a literature-supported design. Separately, product quality is a biological variable, not only a release criterion: double-stranded RNA and residual template trigger innate immune sensing and suppress translation ⁶, so an impure batch can mimic a negative biological result. What must be measured: OSK-versus-OSKMLN comparison in the same assay, plus per-batch dsRNA content, integrity, capping efficiency, residual DNA and endotoxin.

Critical proof-of-concept question.
Can a reproducible OSK-mRNA formulation produce improvement across several aging-associated and functional readouts while preserving the differentiated identity and normal behavior of aged human cells? If the answer is no, further manufacturing scale-up has little value; if the answer is yes, RNA production itself appears comparatively tractable.

AI Usage

Throughout the development of this homework assignment, generative AI served as a drafting aid. It helped structure and refine complex ideas involving partial cell reprogramming, mRNA technology, and motivational background. Particularly by logically organizing and articulating the ideas and goals alongside the proposed actions. While the AI was used iteratively to enhance clarity and explore different framings, all substantive arguments, critical judgments, and final decisions remain entirely my own. Link to prompts

References