Week 1 HW: Principles and Practices

HW1

Ultra-efficient DNA Synthesis Machine

My research is on designing a DNA synthesis machine that can reduce the cost and time to produce long strands of arbitrary DNA sequences. Right now we’re aiming for megabase strands of DNA but the goal of the project is to eventually get to the Gb range and establish a technology that can scale and improve similar to transistors in Moore’s Law. For context, I work mostly on the mechanical side and hope to learn more about the biochemistry and synthetic biology in general from this class.

Most of today’s DNA synthesis tools and services are good at making short DNA pieces, from hundreds to thousands of bps, but as strands get longer, they become much harder and costlier to produce. As an example, Integrated DNA Technologies sells ~125-3,000 bp fragments for uses like cloning single genes and antibody research, the cost is around $0.07-$0.20 per base pair.

Short and mid-length sequences are still really useful:

• 20–100 bp pieces are used for primers to amplify DNA or guide RNA design

• 1,000–3,000 bp can code for single proteins or small metabolic pathways

• 10,000–100,000 bp pieces could represent entire operons

Once hit the megabase range and beyond, you can do a whole lot more:

• A typical bacterial genome like E. coli is ~1–13Mbp

• Yeast has a genome around 12Mbp long

• A gigabase approaches the size of large eukaryotic chromosomes

• 3.2Gbp is the size of the human genome

The goal is not just to achieve longer strands, but to make long, accurate DNA affordable and reliable, opening the door to lots more possibilities.

Governance/Policy

At the gigabase scale, DNA synthesis moves beyond individual genes and can be used to create entire genomes or chromosomes, which raises new safety, security, and ethical considerations. The main governance challenge for lowering the barrier to creating these large constructs is: how this technology and its outputs be developed and deployed responsibly as its capabilities scale.

Goal

Enable the constructive use of large-scale DNA synthesis while preventing misuse or unintended harm enabled by scale and accessibility.

Governance should aim to prevent malfeasance enabled by longer DNA constructs. This involves:

• Preventing the synthesis of harmful DNA sequences
• Reducing the risk posed by unintentional misuse
• Ensuring that increased automation does not eliminate safety checks and that safety checks adapt to the technology

One way to enable this is through system-level design. The hierarchy of controls encourages eliminating risk before having to rely on training or user intent. By incorporating safety checks, containment options, and traceability mechanisms directly into hardware, safe operation becomes the default outcome. Without governance and training, things can still go wrong, and although accessibility is powerful, there should be limits set to ensure compliance and training. In case something goes wrong, there should be an established line of accountability and transparency with oversight.

Governance Actions

To ensure that large-scale DNA synthesis develops in a safe and constructive way there should be a mix of technical, institutional, and regulatory governance actions.

Hardware-Level Constraints on Maximum Assembly Length

• Purpose Currently, DNA synthesis limits are largely economic and biochemical but they could be enforced by hardware limits

• Design Mechanical or software-enforced limits on assembly length, these would be enforced by governance and added by design choices made by manufacturers and research labs building synthesis platforms.

• Assumptions One assumption is that limiting the strand length meaningfully reduces misuse risk. I don’t know where this limit would be set at but it’s possible that you could get around this or that the set length can still lead to misuse. It is also assumed that this wouldn’t hinder legitimate research.

• Risks of Failure & “Success” As mentioned before, you could find ways to misuse shorter strands or combine shorter strands into longer ones. Or this length could keep legitimate research from occurring successfully.

Sequence Screening and Logging

• Purpose Many DNA synthesis companies perform sequence screening to flag known pathogenic or regulated sequences. If synthesis moves toward in-house and machine-based systems, this screening could still take place or even be extended to logging

• Design Automated sequence screening could occur within each synthesis machine with flags dispersed or hardware/software locks enabled if needed. It might also be wise to log synthesized sequences above a certain size threshold. This can be implemented by academic institutions, commercial developers, and funding agencies. When working in MIT nano, your process must be approved by a committee before you begin and EHS reviews new and acceptable chemicals, this would work the same way.

• Assumptions This assumes that known harmful sequences can be meaningfully identified. And that users will accept limited logging in exchange for access.

• Risks of Failure & “Success” It is possible that screening can miss novel or emergent risks. Or that logging could raise intellectual-property concerns. If “too successful,” logging could discourage exploratory research using these systems.

Tiered Access

• Purpose I think one of the best ways to govern this technology is with a tiered access model where synthesis capabilities scale with demonstrated ability, infrastructure, and oversight.

• Design You could start with basic access for short and mid-length synthesis, with fairly open access to this. Then more advanced capabilities could be unlocked by agreeing to institutional reviews, trainings, and safety approval, along with oversight by universities or national research bodies. It might even be better to have a centralized location with the extra advanced machines with specialized oversight on them though still granting access.

• Assumptions This assumes that governing institutions can fairly evaluate readiness and risk and that training and review improve safety outcomes. It also assumes that access tiers won’t become arbitrary gatekeeping, and won’t come down to the same financial barriers in place now.

• Risks of Failure & “Success” This could disadvantage smaller or less well-funded labs, and “success” might slow innovation if approval processes lag behind technology.

Recommended Governance Approach and Trade-offs

Based on the scoring, I would prioritize a combined approach using sequence screening and logging (Option 2) and tiered access (Option 3). Together, these options provide the strongest balance between preventing misuse, enabling response if something goes wrong, and still enabling legitimate research.

Option 2 performs best in terms of biosecurity and response. As DNA synthesis moves toward in-house, automated, and large-scale systems, maintaining some form of sequence screening becomes increasingly important. Screening and logging help ensure that synthesis at larger scales includes visibility and accountability. It also enables backtracking to find problems or errors, which is important when failures or misuse may not be immediately obvious.

Option 3 complements this by recognizing that not all synthesis capabilities carry the same level of risk. A tiered access model allows safe DNA synthesis to remain relatively open, while reserving the most powerful capabilities for users and institutions with appropriate infrastructure, training, and oversight.

Week 2 Lecture Prep

Homework Questions from Professor Jacobson:

• 1: DNA polymerase has an error rate of 1:10^6. With the 3.2 Gb human genome, that’s ~3,200 errors per replication. Biology fixes this discrepancy using polymerase proofreading and the MutS repair system, which functions likea multi-stage error-correction protocol to drop the effective mutation rate.
• 2: If an average human protein has around 400 amino acids, then redundancy allows for 400!/(20!)^{20 which is on the order of ~10}501 DNA sequences to code for an average protein. But many versions fail because RNA secondary structures physically block ribosomes, or sequences trigger RNA cleavage and codon bias issues.

Homework Questions from Dr. LeProust:

• 1: Phosphoramidite synthesis
• 2: Synthesis hits a wall due to an exponential yield drop. Following the (1 - error rate)^N, curve, the probability of a “perfect” strand decreases with every base added.
• 3: At 2kb, the yield of perfect strands is basically zero. To reach the Gb range, we have to assemble smaller, verified oligos using PCA or Gibson Assembly rather than making them in one shot.

Homework Question from George Church:

• 1: Phenylalanine, Valine, Threonine, Tryptophan, Isoleucine, Methionine, Histidine, Arginine, Leucine, and Lysine
• 2: As a fail-safe, the Jurassic Park lysine contingency is flawed because all animals are already naturally unable to synthesize lysine; they get it from food. An escaped organism would do the same and just find lysine-rich food.

Sources

• https://www.reddit.com/r/JurassicPark/comments/12ib7to/the_lysine_contingency_makes_no_sense_right/
• image credit: Bioninja (https://old-ib.bioninja.com.au/standard-level/topic-3-genetics/32-chromosomes/genome-size.html)

AI Prompts

• “how long is an average human protein?”
• “give me some examples of what certain lengths of DNA can achieve?”

Week 2 HW: DNA Read, Write, and Edit

HW2

Part 1: Benchling & In-silico Gel Art

This week, we made gel electrophoresis art using Lambda phage DNA and ten restriction enzymes. Gel electrophoresis uses a positive charge to pull negatively charged DNA through a conductive gel. Longer strands move slower and shorter strands move faster meaning that different lengths of DNA fragments will appear as different bars in your gel. To use this in an artistic context we take our input Lambda DNA and cut it to different lengths using different restriction enzymes which allows us to have coarse control over where these bars end up and thus we can make art with it. I have decided to really commit to my favorite animal, turtles, this semester and try to have a turtle-inspired theme to all of my projects. In an ideal world this is what I wanted my gel art to look like.

Part 2: Gel Art - Restriction Digests and Gel Electrophoresis

This was my first time going through the whole process of making a gel, but I’ll describe more about the experiment in the lab2 page on my website.

Part 3: DNA Design Challenge

3.1: Choose your protein

For my design challenge homework, I have picked Green Fluorescent Protein (GFP) because I was inspired by the HTGAA website photos, it is great for visualizing gene expression, and I want a second chance at making a turtle this time I want to make a turtle-shaped fluorescence pattern.

Here’s the protein sequence from Uniprot: >sp|P42212|GFP_AEQVI Green fluorescent protein OS=Aequorea victoria OX=6100 GN=GFP PE=1 SV=1 MSKGEELFTGVVPILVELDGDVNGHKFSVSGEGEGDATYGKLTLKFICTTGKLPVPWPTL VTTFSYGVQCFSRYPDHMKQHDFFKSAMPEGYVQERTIFFKDDGNYKTRAEVKFEGDTLV NRIELKGIDFKEDGNILGHKLEYNYNSHNVYIMADKQKNGIKVNFKIRHNIEDGSVQLAD HYQQNTPIGDGPVLLPDNHYLSTQSALSKDPNEKRDHMVLLEFVTAAGITHGMDELYK

3.2: Reverse Translate: Protein (amino acid) sequence to DNA (nucleotide) sequence

I used NCBI to get this reverse translation: >reverse translation of sp|P42212|GFP_AEQVI Green fluorescent protein OS=Aequorea victoria OX=6100 GN=GFP PE=1 SV=1 to a 714 base sequence of most likely codons. atgagcaaaggcgaagaactgtttaccggcgtggtgccgattctggtggaactggatggc gatgtgaacggccataaatttagcgtgagcggcgaaggcgaaggcgatgcgacctatggc aaactgaccctgaaatttatttgcaccaccggcaaactgccggtgccgtggccgaccctg gtgaccacctttagctatggcgtgcagtgctttagccgctatccggatcatatgaaacag catgatttttttaaaagcgcgatgccggaaggctatgtgcaggaacgcaccatttttttt aaagatgatggcaactataaaacccgcgcggaagtgaaatttgaaggcgataccctggtg aaccgcattgaactgaaaggcattgattttaaagaagatggcaacattctgggccataaa ctggaatataactataacagccataacgtgtatattatggcggataaacagaaaaacggc attaaagtgaactttaaaattcgccataacattgaagatggcagcgtgcagctggcggat cattatcagcagaacaccccgattggcgatggcccggtgctgctgccggataaccattat ctgagcacccagagcgcgctgagcaaagatccgaacgaaaaacgcgatcatatggtgctg ctggaatttgtgaccgcggcgggcattacccatggcatggatgaactgtataaa

3.3. Codon optimization

I used a handy website called https://www.novoprolabs.com/tools/codon-optimization to optimize my codon for Escherichia coli (E.coli). I chose this organismbecause it is commonly used in synthetic biology, safe, robust, and easy for people new to the wet lab. Why did I have to optimize it at all? GFP comes from a jellyfish it occurs naturally in that species but not in everything, in order to ensure that it will work well with E.coli I have to pick the codons it prefers (codon bias). This is possible because multiple codons can code for the same amino acid. Codon bias can be a problem if the host organism has low amounts of the matching tRNA. So codon optimization replaces rare codons with preferred codons and can remove unwanted restriction sites. Here’s the optimized sequence: ATGTCTAAAGGCGAAGAACTGTTCACCGGTGTGGTTCCGATCCTGGTCGAGCTGGACGGCGACGTAAACGGCCACAAGT TCTCTGTATCTGGTGAAGGCGAGGGTGATGCAACCTACGGTAAACTGACTCTGAAGTTCATTTGCACTACTGGTAAACT GCCGGTTCCGTGGCCGACTCTGGTCACTACTTTCAGCTACGGTGTACAATGTTTTTCCCGTTACCCGGATCACATGAAG CAGCATGACTTCTTCAAATCTGCTATGCCGGAAGGCTACGTTCAGGAACGCACCATCTTCTTCAAAGACGACGGTAACT ACAAAACTCGCGCTGAGGTTAAGTTTGAAGGCGACACCCTGGTTAATCGTATCGAACTGAAAGGCATTGACTTCAAAGA AGATGGTAACATCCTGGGTCACAAACTGGAATACAACTACAACAGCCATAACGTTTACATCATGGCAGACAAACAGAAA AACGGCATCAAGGTGAACTTCAAAATTCGTCACAATATCGAAGATGGTTCCGTGCAGCTGGCCGATCACTACCAGCAGA ACACTCCGATCGGTGACGGTCCGGTGCTGCTGCCGGACAATCACTATCTGAGCACTCAAAGCGCCCTGAGCAAAGACCC GAACGAAAAACGTGATCACATGGTGCTGCTGGAATTCGTTACCGCGGCAGGCATCACTCACGGCATGGATGAACTGTAT AAA

And stats about what’s changed: CAI before optimization: 0.80 CAI after optimization: 0.83 GC content before optimization: 48.60% GC content after optimization: 49.30%

3.4. You have a sequence! Now what?

Cell-Dependent Expression:

To get the GFP DNA to express in a cell, I would insert the codon-optimized GFP gene into a plasmid and transformed into E. coli. The plasmid includes a promoter that allows the cell to recognize and transcribe the gene. Once inside the bacteria, RNA polymerase transcribes the GFP DNA into mRNA, and ribosomes translate the mRNA into the GFP protein. As the protein folds into its final structure, it begins to fluoresce. In this system, the living cell provides all the machinery needed for transcription and translation.

Cell-Independent Expression:

To get GFP to be produced without living cells, I’d need to mix a cell extract containing ribosomes, enzymes, and tRNAs with the GFP DNA template in a test tube. The extract carries out transcription and translation directly in solution. This allows faster protein production and more control over reaction conditions since there is no need to grow or maintain cells. It is useful for rapid testing of gene designs before moving into full bacterial expression.

Part 4: Prepare a Twist DNA Synthesis Order

I chose a similar protein to the example in class but replaced the sfGFP coding sequence with my regular GFP coding sequence. I wonder what the differences are and if they are different enough to create a pattern with. ANyways here’s my benchling linear map.

Here’s my final plasmid:

Part 5: DNA Read/Write/Edit

5.1 DNA Read

If I could choose DNA to sequence, I would choose synthetic DNA used for digital data storage. DNA data storage is interesting because it treats DNA as away of encoding information, like a hard drive, but at a molecular scale. Sequencing the DNA would allow us to read back the stored data and measure how accurately the system preserves information over time, this seems like an idea straight out of science fiction though it is possible now. To sequence this DNA, I would use Illumina sequencing, a second-generation sequencing technology. It works well for short, synthetic DNA fragments and provides high accuracy at relatively low cost. The input would be the synthetic DNA fragments that encode the information. These fragments wouldneed to be prepared by adding adapter sequences to their ends, amplifying them by PCR, and loading them onto a flow cell. During sequencing, fluorescently labeled nucleotides are incorporated one base at a time, and a camera detects the color signal to determine which base was added. This process converts fluorescence into a digital readout of A, T, C, and G. The output is a large dataset of DNA reads that can be reconstructed into the original digital file.

5.2 DNA Write

If I could synthesize DNA, I would create my turtle-themed GFP expression construct but with multiple colors of fluorescent proteins. I would synthesize codon-optimized Fluorescent Protein genes under the control of a bacterial promoter so that it could be expressed in E. coli. The goal would be to design a system that produces a turtle-shell-like hexagonal fluorescence pattern. This connects synthetic biology with spatial design and pattern formation maybe I could even get a system for animating it and getting the turtle to move. The core sequence would be the coding region, inserted into a plasmid backbone.

To synthesize this DNA, I would use chemical DNA synthesis and fragment assembly. Short DNA oligos are chemically synthesized, assembled into the full gene using overlapping regions, cloned into a plasmid, and sequence-verified. This method is good for constructs around 1 kb, such as GFP. Limitations include cost increasing with length, possible synthesis errors, and longer turnaround times for larger constructs.

5.3 DNA Edit

An interesting DNA edit project would be engineering biological motors or force-generating systems that could act as microscopic actuators. It would be interesting to use the process that proteins such as Kinesin-1 use to convert chemical energy from ATP into mechanical motion along microtubules. By editing the genes that encode these motor proteins, we could potentially tune their speed, force output, or binding properties, creating nanoscale linear motors with maybe even the potential to scale up. Editing their DNA could allow us to design programmable biological actuators for soft robotics, microfluidics, or responsive materials.

To perform this editing, I would use CRISPR-Cas9. I’d design a guide RNA that targets the motor protein gene. Cas9, directed by the guide RNA, cuts the DNA at a specific location. If we want to introduce a modification, such as a mutation that alters motor speed or adds a binding domain, we would also provide a donor DNA template for repair. The inputs include the guide RNA sequence, Cas9, the donor template if inserting changes, and host cells. After editing, the modified cells would express the altered motor protein, and its mechanical properties could be measured experimentally. Limitations include variable editing efficiency and the possibility of off-target edits, but the approach might allow for precise modification of biological force-generating systems.

Sources

• https://www.uniprot.org/uniprotkb/P42212/entry#sequences
• https://www.novoprolabs.com/tools/codon-optimization
• https://www.bioinformatics.org/sms2/rev_trans.html

AI Prompts

• “How do you transcribe DNA and translate a protein with a cell-dependent method?”
• “How do you transcribe DNA and translate a protein with a cell-independent method?”

Week 3 HW: Lab Automation

HW3

Lab Preparation: Opentrons Artwork

This week, we programmed the Opentrons liquid handling robot to create fluorescent protein masterpieces. I was really looking forward to this lab and even did last week homework about expressing GFP in E.Coli. Rather than using the GFP, I found we used a variety of different colors of superfluorescent proteins. Ronan’s webtool [1] made it really easy to visualize a design, and we could even upload images to serve as a template for our designs. I decided to go all in on turtles and make a turtles all the way down image featuring a turtle with a globe for its shell. This was the original image, from my collection of Turtle CADS:

I took the coordinates from this image for each different color and put them into the Colab Python Script. I had to add and change colors to match the ones I used. I originally ran into a problem where I’d tell the robot to fill up the pipette, empty it, then continue attempting to dispense I was able to see this error in the simulation and correct it by adding a loop where, after the pipette was empty (20 dots in my case using a p20 and 1ul droplets) I’d pick up more before continuing. I used the built-in Gemini chatbot to help with this assignment, after explaining the functions I had access to I asked it to use these functions to deposit blue droplets at every coordinate in a list. From there, I used the same functions and procedures to do the rest of my colors. After some troubleshooting, the simulation finally showed me the result I wanted, and the verification also went through.

It was really straightforward to go from my design to the Opentron, at least from my point of view as a student (not sure what magic the TAs and staff work behind the scenes). The machine itself was incredibly precise, not only in movement but also in dispensing. Here’s how my final design ended up: I’ll explain more about the process in my lab3 webpage:

Post-Lab Questions

Bioautomation in the wild [2]

The paper I chose was entitled: “TidyTron: Reducing lab waste using validated wash-and-reuse protocols for common plasticware in Opentrons OT-2 lab robots” authored by Bryant et al. (2023). I chose this paper because it touches on two things I care about, one being minimizing lab waste, and two being washing processes.

In this study, the authors developed an automated system, called TidyTron, that runs on the Opentrons liquid-handling robot and reduces laboratory plastic waste by automating wash-and-reuse protocols for common lab consumables like pipette tips and microplates. Traditionally, many of these are discarded after a single use because of uncertainty about cross-contamination, the authors even state that biotechnology labs generate ~5.5 million tons of plastic waste per year. TidyTron addresses this problem by using the OT-2 to perform consistent and reliable cleaning processes that are tested and validated for effectiveness.

The paper describes how the automated protocols were implemented on the Opentrons to rinse, clean, and sterilize plastics used with DNA solutions or microbial cultures. They evaluated the cleaned materials by measuring residual contamination using colony-forming unit (CFU) counts, quantitative PCR to detect residual DNA, and other metrics to confirm that the robot-washed consumables were equivalent in performance to new plastics. Their results showed that the wash-and-reuse procedures could effectively remove contamination, demonstrating that automation makes reusable workflows both practical and safe. I thought this approach was interesting because it applies automation not to increasing throughput or accuracy, but to sustainability in biological labs. I hope to use a similar system, not for plastic but for metal/more permanent lab equipment, which I’ll build into my DNA dispensing machine.

Bioautomation in my final project

As a mechanical engineer, I am very interested in the bioautomation part of synthetic biology. Two of my three final project ideas either seek to improve or apply bioautomation as a main part of the project. One idea I had was to do what we did in lab this week but in 3D. Two approaches I imagine are either to embed a 3D model in a brick of clear gel, similar to the lasers that can engrave a model into glass, or to build a 3D model out of gel and “paint” different fluorescent proteins around it all completely automated. I can imagine different challenges based on which version I pursue, but this would be nearly impossible without an automated system.

Sources

• [1] https://opentrons-art.rcdonovan.com/
• [2] Bryant JA Jr, Longmire C, Sridhar S, Janousek S, Kellinger M, Wright RC. TidyTron: Reducing lab waste using validated wash-and-reuse protocols for common plasticware in Opentrons OT-2 lab robots. SLAS Technol. 2024 Apr;29(2):100107. doi: 10.1016/j.slast.2023.08.007. Epub 2023 Sep 9. PMID: 37696493; PMCID: PMC12212179.

AI Prompts

• “Using the functions described in this document, write a loop that deposits blue droplets at every coordinate in a given list”
• “Use the function to refill the pipette after it’s empty, then continue depositing droplets if there are more droplets than the pipette can hold at once”