Week 1 HW: Final Project Ideas

A “Plug-and-Play” Personalized Cancer Vaccine Modular Biomanufacturing Using Virus-Like Particles and Cell-Free Protein Synthesis Submitted by:Keshav Program: B.Tech Biotechnology Assignment:Homework #1 — Biomanufacturing Proposal

The idea in one line

I want to explore whether we could pre-manufacture a universal virus-like particle and then rapidly attach a patient’s own tumor neoantigens to it, so that a personalized cancer vaccine becomes more like assembling a modular product than starting a new manufacturing process from zero for every patient.

1. What would I like to biomanufacture, and why?

One thing that has always interested me about cancer is that even when two people are diagnosed with the same type of cancer, their tumors are not necessarily molecularly identical.

Tumor cells accumulate mutations, and some mutations can create new peptide sequences called neoantigens. Because these sequences are present in the tumor but not normally present in healthy cells, they can potentially act as targets for the immune system.

That makes personalized cancer vaccines an interesting idea. Instead of giving every patient exactly the same antigen, we could sequence a patient’s tumor, identify promising neoantigens, and manufacture a vaccine designed around that particular tumor.

The problem is that this creates a manufacturing nightmare.

Normal biomanufacturing is designed around:

«one product → manufactured many times»

Personalized vaccines move toward:

«one patient → one slightly different product»

That is the part of the problem I find most interesting.

So rather than manufacturing a completely different vaccine from scratch for every person, I would like to make a modular neoantigen vaccine.

My system would have two pieces:

1. A universal virus-like particle (VLP) scaffold that could be manufactured in relatively large batches and stored.

2. A patient-specific neoantigen module that would only be produced after the patient’s tumor sequence was known.

The patient-specific neoantigens would then be attached to the pre-made VLP.

In other words, I am imagining the VLP almost like biological hardware, while the neoantigen sequence is the part that gets customized.

This is not completely my imagination. Recent research has demonstrated “plug-and-display” VLP systems in which neoantigen peptides can be attached to a pre-existing hepatitis B core VLP scaffold.

What I am proposing is therefore not a new cancer treatment claim. I am more interested in the manufacturing question: could automation and modular biomanufacturing shorten the time between identifying a patient’s neoantigens and physically producing their vaccine?

2. What living system would I use?

I would actually divide manufacturing into a standardized part and a personalized part.

Part A — Universal VLP scaffold

For the reusable scaffold, I would use a recombinant microbial expression system to produce hepatitis B core protein (HBc) virus-like particles.

HBc proteins can self-assemble into highly ordered VLP structures.

The VLP would carry a molecular “catcher” that allows a corresponding tagged neoantigen to be attached later.

The important idea is that this scaffold does not have to be redesigned for every patient.

A large batch could therefore be manufactured, purified, quality-controlled and stored in advance.

Part B — Patient-specific neoantigen

For the personalized component, I would use an E. coli-derived cell-free protein synthesis system (CFPS).

This is the part I find especially interesting.

Normally we genetically engineer a cell and then convince that cell to manufacture our protein. In cell-free expression, we take the molecular machinery responsible for transcription and translation out of the cell and use it directly.

A simplified reaction would contain:

• E. coli cell extract containing ribosomes and translation machinery
• amino acids
• nucleotides
• an energy source
• salts and cofactors
• DNA encoding the selected neoantigen sequence

Then:

DNA template → transcription → translation → neoantigen-containing protein

Because there is no requirement to grow a new engineered cell line for every patient’s sequence, changing the product could be as simple, conceptually, as changing the DNA template.

That is exactly what makes cell-free systems attractive to me for personalized manufacturing.

Modern E. coli cell-free systems have reported recombinant-protein production in the mg/mL range, although the actual yield depends heavily on the protein being produced.

After production, my neoantigen module would contain the complementary molecular tag required to attach it to the universal VLP.

The workflow becomes:

Tumor sequence

↓

Neoantigen selected

↓

DNA template generated

↓

Cell-free production of personalized antigen

↓

Purification

↓

Attach antigen to pre-manufactured VLP

↓

Personalized VLP vaccine candidate

So the slow part of vaccine manufacturing is no longer completely repeated for every patient.

3. Where would automation actually help?

This is probably the part of the project where automation matters most.

I would not automate something simply because pipetting robots exist. The real reason for using automation here is that personalization creates a large number of small, different manufacturing jobs.

Imagine 100 patients.

Instead of manufacturing one protein 100 times, we may need to manufacture 100 different antigen sequences.

Doing that manually would become difficult very quickly.

My automated workflow

After sequencing and computational neoantigen selection, a robotic liquid-handling system such as Opentrons could receive the DNA templates belonging to different patients.

The robot could prepare parallel cell-free reactions containing the same master mix but a different DNA template in each reaction.

For example:

Reaction| DNA template Patient 1| Neoantigen set 1 Patient 2| Neoantigen set 2 Patient 3| Neoantigen set 3 …| … Patient 96| Neoantigen set 96

The robot could then automate:

• reaction setup,
• reagent dispensing,
• incubation setup,
• sampling,
• purification preparation,
• concentration measurements,
• VLP–neoantigen coupling reactions,
• and preparation of analytical assays.

A plate reader or other analytical instrument could return the measurements to the same workflow.

Eventually I would like the system to behave almost like a design-build-test loop:

sequence → manufacture → measure → accept/reject

rather than having a researcher manually coordinate every individual sample.

This is also where a Cloud Lab becomes interesting. A computational pipeline could generate the design, while standardized robotic equipment executes the physical workflow.

The important advantage is therefore not simply speed.

It is traceability and reproducibility when every sample is different.

4. How much would I need to manufacture?

For the back-of-the-envelope calculation, I will imagine a small personalized-vaccine program treating:

100 patients

Existing personalized neoantigen vaccine trials have used doses containing roughly hundreds of micrograms to a few milligrams of total neoantigen peptide.

For example, the EVX-01 clinical study evaluated total peptide doses from approximately 0.5–2 mg per vaccination.

I am not claiming that this dose transfers directly to my VLP system, because the immunogenicity and antigen density would be different.

For an engineering calculation, however, I need a starting assumption.

I will assume:

0.5 mg personalized neoantigen material / dose

and

6 doses / patient

Therefore:

0.5 mg/dose × 6 doses

= 3 mg per patient

For 100 patients:

3 mg/patient × 100 patients

= 300 mg personalized antigen

I will add a 20% manufacturing allowance for purification losses, analytical testing and material that may fail QC.

300 mg × 1.20

= 360 mg

So my manufacturing target is approximately:

360 mg of usable patient-specific antigen

The important difference from conventional manufacturing is that this is not one 360-mg batch.

It is effectively:

100 different ~3.6-mg batches.

That changes the engineering problem completely.

5. How much cell-free reaction would I need?

Published cell-free systems can reach protein yields in the mg/mL range, but using the highest published yield for my calculation would probably be overly optimistic.

So I will assume a more conservative usable yield of:

0.5 mg/mL

after taking expression and recovery losses into account.

For one patient:

3.6 mg ÷ 0.5 mg/mL

= 7.2 mL reaction volume

I would round this up to approximately:

10 mL CFPS reaction per patient

For 100 patients:

10 mL × 100

= 1,000 mL

or

approximately 1 L total cell-free reaction volume

Again, the interesting part is that I would not necessarily run one 1-L reactor.

I would prefer many parallel small reactors, because every patient requires a different DNA template.

For example:

100 × 10-mL reactions running in parallel

would make more sense for this system than one conventional fermenter.

That is almost the opposite of classical industrial biotechnology, where we normally try to make reactors larger.

6. How long would manufacturing take?

This is where I would hope the modular system provides its main advantage.

Current personalized peptide-vaccine manufacturing can take several weeks. In one published personalized vaccine study, manufacturing took approximately 48–55 days.

I would therefore set myself a much more aggressive engineering target:

less than 5 days from an approved neoantigen sequence to a finished personalized antigen-VLP candidate.

A rough workflow could be:

Day 0 Patient-specific DNA templates become available.

Day 1 Cell-free expression of the personalized antigen.

Cell-free reactions can produce recombinant protein within hours rather than requiring several days of cell cultivation.

Day 2 Purification and concentration measurement.

Day 3 Couple personalized neoantigens to the pre-manufactured VLP scaffold.

Day 4 Analytical characterization and initial QC.

So my estimated manufacturing time is approximately:

4–5 days

after the antigen sequence and DNA template are available.

I want to be careful here: this does not mean a patient could actually receive an approved vaccine four days after a biopsy.

Tumor sequencing, neoantigen prediction, sterile manufacturing, extensive quality testing and regulatory release would all add time.

The 4–5 day estimate refers only to the manufacturing module that I am proposing to redesign.

7. What happens when this becomes larger?

Interestingly, I would not necessarily scale this process by continuously increasing reactor size.

I would scale it by parallelization.

For example:

100 patients

→ 100 parallel reactions

1,000 patients

→ multiple automated manufacturing modules

10,000 patients

→ distributed automated facilities using the same standardized platform

This is similar to the difference between making one enormous computer and manufacturing many processors in parallel.

The universal VLP scaffold could still be produced using conventional larger-scale fermentation because it is identical for everyone.

Only the small personalized module would use distributed manufacturing.

So my system intentionally combines two manufacturing philosophies:

mass manufacturing for the universal component

automated distributed manufacturing for the personalized component

I think this hybrid model makes more sense than trying to personalize every part of the process.

8. What don’t I know yet?

This is also where the idea becomes much less clean than it looks on paper.

1. Will my cell-free yield actually be high enough?

I used 0.5 mg/mL as an engineering assumption.

But cell-free expression is very protein-dependent.

Some neoantigen constructs may express beautifully while another patient’s sequence may aggregate or barely express at all.

That means the biggest question may not be:

«“What is the average yield?”»

but:

«“How variable is the yield from patient to patient?”»

For personalized manufacturing, variability might be more dangerous than low average productivity.

2. How many displayed neoantigens are actually needed?

Attaching protein to a VLP does not automatically create an effective cancer vaccine.

I would need to determine:

• how many copies of each neoantigen are displayed,
• whether they remain accessible,
• whether the VLP stays structurally intact,
• and whether antigen-presenting cells process the displayed neoantigens correctly.

So one major optimization variable would be antigen density on the VLP surface.

3. Can quality control be made as fast as manufacturing?

I think this may actually become the biggest bottleneck.

If I can manufacture the protein in one day but characterization takes two weeks, then speeding up expression has not solved the real problem.

For every personalized batch I would need confidence in things such as:

• identity,
• purity,
• concentration,
• VLP assembly,
• neoantigen attachment,
• sterility,
• and batch-to-batch reproducibility.

This makes me think that automated analytical testing may ultimately be just as important as automated manufacturing.

Final thought

When I started thinking about this problem, my first instinct was simply to choose a recombinant vaccine and ask how to produce more of it.

But personalized medicine changes the question.

For a personalized cancer vaccine, we may not need to manufacture an enormous quantity of one molecule. Instead, we need to manufacture small quantities of many different molecules, quickly and reliably.

That is why I chose a modular VLP + cell-free system.

The part I would manufacture in bulk stays constant.

The part that makes the therapy personal is manufactured only when needed.

And automation connects the two.

For me, that is a more interesting version of biomanufacturing than simply asking how large a fermenter needs to be.

References

1. Virus-Like Particle-Based Personalized Neoantigen Nano-Vaccine for Tumor Immunotherapy and Recurrence Prevention. ACS Nano (2025), 19(40), 35385–35400. DOI: 10.1021/acsnano.5c06278.

2. Karim, A. S. & Jewett, M. C. (2021). Toward sustainable, cell-free biomanufacturing. Current Opinion in Biotechnology, 69, 136–144.

3. Petzold, C. J. & Mukhopadhyay, A. (2025). From bench to biofactory: High-throughput technologies and automated workflows to accelerate biomanufacturing. Current Opinion in Biotechnology, 94, 103320.

4. Mørk, S. K. et al. (2024). Dose escalation study of a personalized peptide-based neoantigen vaccine (EVX-01) in patients with metastatic melanoma. Journal for ImmunoTherapy of Cancer.

5. Rao, N. H. et al. (2011). Expression of codon optimized major capsid protein of human papillomavirus in Pichia pastoris: purification and characterization of virus-like particles. Vaccine, 29, 7326–7334.

6. Asin-Garcia, E. et al. (2025). A snapshot of biomanufacturing and the need for enabling research infrastructure. Trends in Biotechnology, 43(5), 1000–1014.

AI Use

AI client: OpenAI ChatGPT, GPT-5.6 Sol

How I used it: I used ChatGPT to brainstorm a biomanufacturing idea connected to my interests in vaccines and cancer biology, identify literature relevant to modular VLP and cell-free manufacturing, check the order-of-magnitude calculations, and help organize my ideas according to the homework questions.

Final Project