Week 1 HW: Principles and Practices

Systemic Lupus Erythematosus (SLE) disproportionately affects women at a 9:1 female/male ratio (McDonald et al., 2015), with higher prevalence in Latin American populations compared to European ancestry groups (Ugarte-Gil et al., 2023). In Ecuador, qualitative studies document significant barriers to SLE treatment and care for women in rural and Andean Indigenous communities, where cultural mismatches with Western medicine delay diagnosis and treatment (Miles, 2011; Bautista-Valarezo et al., 2021). Current immunosuppressive therapies require frequent clinical monitoring unavailable in these settings, leading to uncontrolled disease activity and organ damage.

I want to propose the design of an in vitro prototype of a synthetic mammalian gene circuit. The idea is that the circuit autonomously senses and regulates type I interferon (IFN-α) levels—a validated biomarker of poor outcomes- elevated in over 70% of systemic lupus erythematosus (SLE) adult patients (Whittall Garcia et al., 2024). When IL-10 is chronically exposed to IFN-α in vivo, it becomes part of a pro-inflammatory signaling, instead of its original anti-inflammatory function (Yuan et al., 2011). So, by detecting and regulating IFN- α levels, it will be possible to restore IL-10 anti-inflammatory properties and improve patients’ outcomes. Communities already distrust the traditional healthcare system, so maybe they would be more receptive to this approach, that for them may seem less invasive and restrictive if conducted correctly.

A primary governance goal necesary for developing the project would be to: Ensure bioethical, development of autonomous therapies for diseases affecting marginalized women.

To reach it, I consider the following subgoals:

1. Bioethical guided protocol and trial design: Before human trials, replicability and circuit stability should be guaranteed at least in cellular lines. Bioethics are of utmost importance when designing a biomedical protocol, even more when treating vulnerable groups (e.g. indigenous women). All research decisions are to be made taking into consideration patients’: a) direct personal gains/ benefits from the trial; b) informed consent and willingness to be involved in the trial; and c) rights and personal safety.
2. Culturally grounded consent: Develop dynamic consent frameworks respecting communal decision-making models in Indigenous Kichwa communities (Bautista-Valarezo et al., 2021).
3. Rural accessibility: Guarantee technology pathways reach women in communities >2 hours from tertiary hospitals.
4. Prevent therapeutic abandonment: Ensure patients receive ongoing support if circuit recalibration is needed.

Analyzing the scope of the project, derivated three ordenance actions that meet real concerns came to mind:

To implement the three ordenance actions, they were evaluated in function of how they follow the proposed Governance’s criterion (from 1 to 3, with 1 as the best):

Based on the scoring framework above, I would prioritize a sequential implementation strategy: Option 1 (Bioethical Commitment) and Option 2 (Community Co-Design) as mandatory requirements during the research and early clinical phases, with Option 3 (Access Democrtization) introduced only after successful Phase II clinical validation. This sequencing addresses the fundamental trade-off between safety and accessibility—while open-source democratization (Option 3) scores highest for rural accessibility and minimizing economic barriers, its weak performance in patient safety (score 3) makes premature implementation ethically unacceptable for an autonomous therapy affecting vulnerable indigenous women. Containment requirements (Option 1) provide the non-negotiable safety foundation (scores 1,1 for replicability and patient safety), while co-design (Option 2) ensures cultural legitimacy by centering communal decision-making models (scores 1,1 for culturally-grounded consent)—a critical factor given documented Indigenous distrust of medical systems that dismiss ancestral knowledge (Bautista-Valarezo et al., 2021). The primary trade-off I acknowledge is timeline extension: mandatory co-design workshops and dual-redundant kill switch validation will delay deployment compared to a purely technical development path. However, this delay prevents two greater harms: (a) therapeutic abandonment if communities reject culturally inappropriate designs, and (b) catastrophic safety failures from insufficient containment.

My recommendation targets three audiences simultaneously: (1) locally, the Ethics Committees (CEI) of hospitals collaborating with universities’ CEI should adopt co-design requirements immediately; (2) nationally, Ecuador’s Ministry of Public Health (MSP) must develop tiered containment guidelines specific to synthetic biology therapies, preferably within a set time frame; and (3) internationally, the WHO Ethics Advisory Committee should enforce global standards recognizing that “informed consent” for autonomous therapies must accommodate communal decision-making structures in Indigenous contexts. This phased approach honors both the precautionary principle and justice: safety first, then equitable access, never the reverse.

REFLECTING ON WEEK 1,

There was a new ethical concern (for me) that emerged: The tension between the developing circuit regulated therapies for a patient, and the status quo in decision-making embedded in family/community networks that may forbid the patient access to treatment.
In Ecuadorian indigenous communities, health decisions often involve extended family consultation, especially when there is a preexistent prejudice with the local public system. An autonomous circuit designed based on individualism might undermine culturally valued collective decision-making.
As stated before, this requires governance actions that bring together key members first, to reach the whole community. On a more detailed note, circuit designers can reach a middle ground with the indigenous by showcasing how their system interfaces can be linked with existing social decision-making structures. For example: Could the circuit include a “real-time feedback” function to let family members know if an input (e.g. related to their traditional medical care system, or whether to activate a kill switch) could aid the patient? With this I have learned that deployment strategy is as important as technology and funding to effectively reach target communities.
This exercise revealed that what we perceive as autonomy is not a universal value but culturally constructed. Ethical synthetic biology requires not just technical safety, but cultural open mindedness—designing with communities rather than for them. As future builders of biological systems, we must ask not only “Can we build this?” but “Should we build it this way, for these people, in this context?”

Homework Questions from Professor Jacobson:

Nature’s machinery for copying DNA is called polymerase. What is the error rate of polymerase? How does this compare to the length of the human genome. How does biology deal with that discrepancy? The slides (p.8 & p.10) state that polymerase has an error rate of ~1:10⁶ (1 mistake per million base pairs) and that the human genome length has a length of 3.2 Gbp (which could be expressed as 3.2 × 10⁹ bp). Thus, in humans, each replication generates approximately 3,200 mistakes. Biology –not synthetic biology- copes with this by relying on 3’→5’ proofreading, a mechanism where polymerase detects incorrect base pairings, retires the wrong nucleotide and replaces it as needed.

How many different ways are there to code (DNA nucleotide code) for an average human protein? In practice what are some of the reasons that all of these different codes don’t work to code for the protein of interest? According to slide 6, the average human protein length is 1036 bp. A codon (a base pair triplets) codes for one protein. Assuming that the given protein length on the slide excluded non-coding base pairs of the gene, we have: 1036/ 3= ~345 amino acids. The protein length is fixedly of ~345 amino acids, however due to codon degeneracy, more than one base pair triplet combination/codon can code for the same amino acid (e.g. both AGA and AGG code for Arg). So, with 3 codons per amino acid and 345 amino acid “slots” an average human protein can have: 3³⁴⁵ possible coding ways. In nature, a huge percentage of these coding ways are not viable, as there exists a codon usage bias, set by natural selection, to improve efficiency and precision in translation. The slides (p.39- p. 43) show that regions with a high content of GC (and AT) pairs form hairpin-like structures. These interfere with transcription/ translation as they can act as signaling regions, inducing splicing or stopping transcription/translation. Thus, maybe codon usage bias intents to keep G-C % to a minimum.

Homework Questions from Dr. LeProust:

What’s the most commonly used method for oligo synthesis currently? Why is it difficult to make oligos longer than 200nt via direct synthesis? Why can’t you make a 2000bp gene via direct oligo synthesis? In the slides (p.11) the first automated method of oligo synthesis was, proposed by ABI (Applied Biosystems), with subsequent optimizations. However, it was mainly based on the CPG method, which in turn uses the phosphoramidite method, both designed by Caruthers (Matteucci & Caruthers, 1981). So, at its core, the most used is the phosphoramidite method. The cycle consists of phosphoramidite-based synthesis consists of four steps repeated per nucleotide addition: -Coupling: Phosphoramidite addition -Capping: Blocking unreacted sites -Oxidation: Stabilizing the bond -Deblock: Removing the protecting group for the next cycle The yield per cycle is determined by the coupling efficiency and the length (in bases) of the oligonucleotide by the formula:

So, even if coupling efficiency always reaches a 90.0%-99.5% range, the longer the length, the poorer the yield (Merck, 2026). This limitation is shown when applying the formula to an oligo with a length of 200 bases at 99,5%: Yield= 0.995200 ^ 200= 0.367= 36.7% The formula works as it explains that in each cycle, while 99% of the oligonucleotide chains are synthetized, the remaining 1% is truncated and starts to accumulate. Only 0.995 ^ (n cycles) of the original chains reach complete length. 
Based on that logic, if a direct oligo synthesis was performed for an oligo with a 2000 bp length, the yield would not be significative (<0.001%). On the other hand, by following the multiplexed gene fragment solution proposed on the slides, the overall yield increases, making synthesis more accurate as a whole.

Homework Question from George Church:

Given slides #2 & 4 (AA:NA and NA:NA codes), what code would you suggest for AA:AA interactions? Nucleic acid/Nucleic acid (NA:NA) code refers to the nucleotides forming determined/ fixed base pairs that will always bond with one another: “A” bonds with “T” (“U” in RNA), and “C” bonds with “G”. Based on this empirical rule, this type of code becomes a programable basepair code, allowing us to design a sequence with certainty. In the case of Amino acid/ Nucleic acid (AA:NA) code, the rule is that the ribosome can translate one stranded sequence - by reading codons, base triplets- into amino acids. Each codon codes for a specific amino acid, allowing the synthesis of bigger proteins. However, in translation, there is more than one codon that can be translated into the same amino acid (codon degeneracy). In that line, the amino acid/ amino acid (NA: NA) code should be the set, deterministic rule that describes all amino acid interactions. Protein primary structures result from a covalent bond between the N terminus of one to the c terminus of another. Nevertheless, there is not such a thing as a generalized fixed way of interaction among amino acids, due to the presence of their variable side chains. The functional groups make the side chains unique, directly giving amino acid their properties: polarity, charge/ attraction (covalent bonds) and the ability to form hydrogen bridges. When interacting, those properties will make interactions-that cause secondary, tertiary and quaternary structures- no deterministic, but statistical and contextual. So, the difficulty of assigning a code to amino acid interactions lies mainly in the variability introduced by the side chain. And that same variability what makes them form functional proteins. Bertozzi biorthogonal reactions occur independently without affecting the natural biochemistry/natural reactions of molecules in the cell. In them a “clic” reaction occurs between an azide and an alkyne cycle, resulting in in a new compound that is a union of both compounds, like “connecting to parts of a seatbelt” (Agard et al., 2005). Now this technology is widely used for fluorescent marking and observation of cellular processes. But maybe we could go an step further, and directly explore how modifying side chains to have azide and alkyne cycles as functional groups impact in improving predictability of secondary, tertiary and quaternary structures.

REFERENCES

• Agard, N. J., Prescher, J. A., & Bertozzi, C. R. (2005). A strain-promoted [3 + 2] Azide−Alkyne cycloaddition for covalent modification of biomolecules in living systems [J. am. Chem. Soc. 2004, 126, 15046−15047]. Journal of the American Chemical Society, 127(31), 11196–11196. https://doi.org/10.1021/ja059912x
• Bautista-Valarezo, E., Duque, V., Verhoeven, V., Mejia Chicaiza, J., Hendrickx, K., Maldonado-Rengel, R., & Michels, N. R. M. (2021). Perceptions of Ecuadorian indigenous healers on their relationship with the formal health care system: barriers and opportunities. BMC Complementary Medicine and Therapies, 21(1), 65. https://doi.org/10.1186/s12906-021-03234-0
• Matteucci, M. D., & Caruthers, M. H. (1981). Synthesis of deoxyoligonucleotides on a polymer support. Journal of the American Chemical Society, 103(11), 3185–3191. https://doi.org/10.1021/ja00401a041
• McDonald, G., Cabal, N., Vannier, A., Umiker, B., Yin, R. H., Orjalo, A. V., Jr, Johansson, H. E., Han, J.-H., & Imanishi-Kari, T. (2015). Female bias in systemic lupus erythematosus is associated with the differential expression of X-linked toll-like receptor 8. Frontiers in Immunology, 6, 457. https://doi.org/10.3389/fimmu.2015.00457
• Merck. (2026). Polymerase Chain Reaction ApplicationsDNA Oligonucleotide Synthesis DNA Oligonucleotide Synthesis. Sigmaaldrich.com. https://www.sigmaaldrich.com/EC/en/technical-documents/technical-article/genomics/pcr/dna-oligonucleotide-synthesis
• Miles, A. (2011). Emerging chronic illness: women and lupus in Ecuador. Health Care for Women International, 32(8), 651–668. https://doi.org/10.1080/07399332.2010.529214
• Ugarte-Gil, M. F., Fuentes-Silva, Y., Pimentel-Quiroz, V. R., Pons-Estel, G. J., Quintana, R., Pons-Estel, B. A., & Alarcón, G. S. (2022). Global excellence in rheumatology in Latin America: The case of systemic lupus erythematosus. Frontiers in Medicine, 9, 988191. https://doi.org/10.3389/fmed.2022.988191
• Whittall Garcia, L. P., Gladman, D. D., Urowitz, M., Bonilla, D., Schneider, R., Touma, Z., & Wither, J. (2024). Interferon-α as a biomarker to predict renal outcomes in lupus nephritis. Lupus Science & Medicine, 11(2). https://doi.org/10.1136/lupus-2024-001347
• Yuan, W., DiMartino, S. J., Redecha, P. B., Ivashkiv, L. B., & Salmon, J. E. (2011). Systemic lupus erythematosus monocytes are less responsive to interleukin-10 in the presence of immune complexes. Arthritis and Rheumatism, 63(1), 212–218. https://doi.org/10.1002/art.30083

Week 2 — DNA Read, Write, & Edit

Part 0: Basics of Gel Electrophoresis

I reviewed this week’s lecture and recitation videos, which provide excellent foundational explanations of gel electrophoresis principles. Having hands-on experience running agarose gels in my research lab, I can confirm that all the technical considerations mentioned are crucial for successful results. One safety tip that particularly resonated with me was handling hot flasks after microwave use. In our lab, we always used heat-resistant gloves, yet we were still advised to minimize contact time with hot containers—a precaution I now understand is essential.

Part 1: Benchling & In-silico Gel Art

I began by locating the Lambda DNA identifier in NCBI’s Genome database. My search strategy involved the keywords “bacteriophage lambda” and the known genome length (48,502 bp). The search returned three results; I selected the first entry because its Assembly status indicated “Viral Project,” which typically signifies a complete genome sequencing effort. Upon verification, the genome size matched the expected 48.5 kb. From this entry, I extracted the GenBank identifier (J02459.1) and directly imported the sequence into Benchling via the “DNA/RNA sequence” import function.

Considering the genome’s 48.5 kb size, I initially planned to use a 50 kb ladder for reference. However, after performing a test restriction digest with EcoRV, HindIII, SacI, KpnI, and SalI, Benchling’s fragment table revealed that all resulting fragments were under 5 kb.

Consequently, I switched to a 5 kb ladder (Geneon’s “DNA Ladder mix - high range”) for my gel art design. The final pattern aims to form a heart shape, as visualized below:

Part 3: DNA Design Challenge

Building on my Week 1 project, I focused on modulating IFN-α levels to reduce poor outcomes in systemic lupus erythematosus (SLE) patients. Current pharmacological approaches for severe SLE also target IFN-α inhibition (Baker et al., 2024). Literature indicates that IFN-α activation occurs through a signaling cascade initiated by Interferon Regulatory Factor 7 (IRF-7). When IRF-7 is no longer needed, the body naturally deactivates it via TRIM28, a SUMO E3 ligase with specific activity against IRF-7 (without affecting the similar IRF-3) (Liang et al., 2011). This mechanism offers a more targeted, less invasive therapeutic approach for SLE.

I retrieved the TRIM28 amino acid sequence from NCBI Protein using the search terms “TRIM28” and “Homo sapiens.” From the results, I selected the third hit (the first two were fragments), obtaining the complete protein sequence in FASTA format: “>AAH52986.1 TRIM28 protein (Homo sapiens) MAASAAAASAAAASAASGSPGPGEGSAGGEKRSTAPSAAASASASAAASSPAGGGAEALELLEHCGVCRERLRPEREPRLLPCLHSACSACLGPAAPAAANSSGDGGAAGDGTGPAKSRDGERTVYCNVHKHEPLVLFCESCDTLTCRDCQLNAHKDHQYQFLEDAVRNQRKLLASLVKRLGDKHATLQKSTKEVRSSIRQVSDVQKRVQVDVKMAILQIMKELNKRGRVLVNDAQKVTEGQQERLERQHWTMTKIQKHQEHILRFASWALESDNNTALLLSKKLIYFQLHRALKMIVDPVEPHGEMKFQWDLNAWTKSAEAFGKIVAERPGTNSTGPAPMAPPRAPGPLSKQGSGSSQPMEVQEGYGFGSGDDPYSSAEPHVSGVKRSRSGEGEVSGLMRKVPRVSLERLDLDLTADSQPPVFKVFPGSTTEDYNLIVIERGAAAAATGQPGTAPAGTPGAPPLAGMAIVKEEETEAAIGAPPTATEGPETKPVLMALAEGPGAEGPRLASPSGSTSSGLEVVAPEGTSAPGGGPGTLDDSATICRVCQKPGDLVMCNQCEFCFHLDCHLPALQDVPGEEWSCSLCHVLPDLKEEDGSLSLDGADSTGVVAKLSPANQRKCERVLLALFCHEPCRPLHQLATDSTFSLDQPGGTLDLTLIRARLQEKLSPPYSSPQEFAQDVGRMFKQFNKLTEDKADVQSIIGLQRFFETRMNEAFGDTKFSAVLVEPPPMSLPGAGLSSQELSGGPGDGP”

Next, I used NCBI’s “tblastn” BLAST tool to identify the corresponding nucleotide sequence. The top hit (Accession NG_046945.1) showed the highest Query Score (82%) and lowest E-value (3e-164), though the alignment revealed significant gaps between amino acid and nucleotide sequences.

Since multiple codons can encode the same amino acid, nucleotide sequences vary across taxa even when the protein sequence remains identical. Therefore, codon optimization must align with the intended host organism. Translation efficiency depends on tRNA abundance; optimization replaces rare codons with synonymous frequent ones matching the host’s tRNA pool. For my project’s therapeutic goal, I optimized the sequence for human cells.

(Note: The optimization tool flagged this sequence as “high complexity,” indicating potential synthesis challenges with commercial methods.)

With the human-optimized sequence ready, the next step is transforming this digital information into a functional protein capable of immune modulation. Two feasible routes exist: cell-dependent or cell-free systems. I could use a cell-free system for initial circuit prototyping (better cost-benefit ratio) and transition to live-cell systems for actual production.

Part 4: Prepare a Twist DNA Synthesis Order

For this exercise, I designed a vector for constitutive TRIM28 expression. In the final project version, I plan to add an IFN-α sensor that inhibits TRIM28 production when concentration exceeds a specific threshold. For now, expressing TRIM28 alone in human cells requires these regulatory elements (McCarty et al., 2023): CMV promoter, Kozak sequence (5’ UTR), Start codon, Optimized TRIM28 coding sequence, His Tag, Stop codon and SV40 polyadenylation signal (3’ UTR).These elements ensure transcriptional and translational efficiency and accuracy.

In Benchling, starting from an empty file:

• I sequentially assembled these components to form the cassette:

1. CMV Promoter:

CGTTACATAACTTACGGTAAATGGCCCGCCTGGCTGACCGCCCAACGACCCCCGCCCATTGACGTCAATAATGACGTATGTTCCCATAGTAACGCCAATAGGGACTTTCCATTGACGTCAATGGGTGGAGTATTTACGGTAAACTGCCCACTTGGCAGTACATCAAGTGTATCATATGCCAAGTACGCCCCCTATTGACGTCAATGACGGTAAATGGCCCGCCTGGCATTATGCCCAGTACATGACCTTATGGGACTTTCCTACTTGGCAGTACATCTACGTATTAGTCATCGCTATTACCATGGTGATGCGGTTTTGGCAGTACATCAATGGGCGTGGATAGCGGTTTGACTCACGGGGATTTCCAAGTCTCCACCCCATTGACGTCAATGGGAGTTTGTTTTGGCACCAAAATCAACGGGACTTTCCAAAATGTCGTAACAACTCCGCCCCATTGACGCAAATGGGCGGTAGGCGTGTACGGTGGGAGGTCTATATAAGCAGAGCT

2. Kozak Sequence: GCCGCCACCAUGG
3. Start Codon: ATG
4. Human-optimized TRIM28 sequence:

TGT GTC CTG CCA TGG CCC CAC CCT CTG CAC TTG CCT AAC AAG TGG CCC TCT GTT ACC GAG CAG GCG GGA AAC GAG GCC CCC ACT CAT TCT GCT AGT TGA GTC GGG GGC GGC GAG GGC CAA GGC CCT GCC CTT TGA TTT CCA CTC CCT AGA GCT TCT CAG GTT TCT GGC TGG TAG ACC TAT TGT CTT CTG CAA AGA ACA CAA GCA TAG ACA CCT TGC GCC GTG CTC TAA GAG CTT TAA TAT AGC CAC CTG CCC AGG CTC CCT GCA CAG TGT CCC CAG GGG CCT CCA GTC TCT CCC AAG GCC TAA TGG CTG GGC GGC TGG GTC CCA CCT CTG CCC TAA CTC AGC GCT CAG GAG CAC ATA TGC CTG CTG AGC GGA ACT TCT AGC TGA AGG ATG CAA TGA GGC ACC AGC GCA TCC TCT TGG CCC CAC TGG TAG TCA GCG CTG GGG ACG AAC ATG CAG CAC TGT CGA CGC GCA CCC AGG CGC TTT GCA GCA CGC AAG TGT GGA TTC TGG GGA TGC GGG GGG GGT CCC GGT CAG CAA GAT CCG ACG TGA CCT GAC CTG CTG TGC CCC CTG GAA AGT GCA AGA TGC CTG ACA TAT AGA TCT GTA TGC AAG TGG ATG TCA CGG TGG CCG TCC TGT CGG TCC TAA CGG TCC TAG ATC AGC GGT GCC GTG TGC TGG TCC ATG ATG CCA CGC TAA GCA TTG TGC AGA TAG GAG GGG TCC CTG TCC CCC CTG TGT TAA CTG ATG ATG CTC AGT GGA GTG CGA AGC GAC CCA AGC ATC CTG CTT CCA AGC CTG CTC ACT TCA TAT ACT TTT TAC CTG CAG AAA GTC ACC GAG GGA CAA CAG GAG CGC CTG GAA AGA CAA CAC TGG ACA ATG ACG AAA ATT CAG AAG CAT CAG GAA CAC ATT CTC CGC TTC GCG TCT TGG GCG CTG GAA AGC GAT AAC AAC ACG GCA CTG TTG CTG AGC AAA AAA CTG GTC TGC ACT GGA GGC TTG CTG GCA GGC GGC TTC CAG GCT GGG GGC AGC CAG TAC CTT AAA AGT CTG TTT GTC GTC GGG GTC CTC ATC TTG TCA TCT CTG AGC CTG CCC ACG AGG TCC ACT AGT AGC TGT ACC GGC CCT AGC CGC TGA CTG TGG ATA CCA TGG AGC CAC ATG GCA AGG TGA AGT TTT TCC GGA ACT TCC ATG CCT GGC CCG AGG GTG CCT AGA CCG CTG GTG GGC CCT CAG CTC TAT TTG ACT CTG TTG CTG CCA CAT GTT CTG CAT CTG CTG ACC TTT CTG TCT AGT TTC TGC CGC CAG GAT TGT GGC CGC GCT AGT TGG CAC TGA CTG AAC AGA CCC TGT ACC CAT GGC CCT TCT AAG AGC CCC AGA GCT CCG GAA CAG GCC GGG CTC TGG CAG CAA CCC GGC GAG CAG GAG CGG GGC CCA CGC AAA GGG TGG GCG GGC AAT GGA GTT CAA TAG CAG GAG AGG GGA CCA CGC CAG GGA GGC TGG GCA GGT AAC GGC GTA CAG TAG GGA AGT TGT CCT ACA GAG GCT GAG GGG AGC GCA CTG TTC ACC GCA CAC GGC GGA GCA GGG AGA CTC TGG CTT TGG GTG CGC TAA GTA GGC CTT CCC TCA GGT TGG GGT AGA GCC CCC GTG TTG TTC CTG CCA CAG CCC GCC AGC CTG CTT TCC CCA TTC TCA AAA GAG ATG ATC CCG ACC CAG GTA CAG TCT CCC ATG TGT CAG GTG TAG AAC GGG AAG TAC GGC ACG TCC CCA GGC GGG GAG GTC GAT GGC GGT TGG GGT TGT ACC CAT GCA AGA CTC CTT CAT TTG CAG GTT CCC CTT CGC TGA GGG CGG GGA GAA CGC CCG TAT GCC CAA GGC GCA ACT TCC GAG CCA TGA ACG CCA GGC CCT GGC CCC CAC TCT TAA CAG CCC GCA ACA AGG CTG CAG GGT CTC CCA AGG CAG TAC CAC TGA GGC CTG CAG CCC TAC TGC TAC TAG ACG TGG CGA TGC CGA TGC TCC TAC CGA CCG GCC CGA GAT TGT GCT TGT AGG AAT CCA TGG TGC CCT ACT CCA GGG TGG CAT GGC CAC TGC CAG GGT AAG CCA GTG CCT CGC AAT TAC AGC TGT CGA ATG AAA CCT GTG GTG CAG GTG TGG ACC TTG CTC AAT ACG CCC CAT CAC CAC CTG GCT GGC GGA GGC GAC GGG GGC TGT CAC TGG TCT CCC TCC TAC TGC CAT TGA GGG CCT TAA GAC CAG ACT TGT GCC TAC GGG AGC TGC GGC GGC TCA TGG TGC TAA GGG TCC CCC CCT GGA CTG ACA TGA TGG CAG CAC CAG CTG AGG GCC GGG GGC GGA GGG TCT TAG GGG TAT CTG AGC CCG CGA TGG TGG CCA GGT AAC CCT GGG TGA CAG TGC CAT CAC CTG CCC TGC CTG CCC GAA GCT CGA CGA TCT GGC TAC GTC CAG CCA GTA TAA GTG CTG TTC CCT CCT GGG CTC TCT CCC GCC GGG CCT GCC GGA TGT ACA CGG GTC AGT GTG CGC TTG GTG GGG GTG AAG AGC GGG TGT TGG GCC GTC TGG ACA GGC TCC TGC GCA GAT CCT TAT TTT CTG CAC CCA AGG GAA GAG TGG TCT TGT AGC CTC TGT CAC GTT TTG CCC GAC CTG AAA GAA GAA GAT GGC TCA CTC TCC CTC GAC GGG GCA GAC TCT ACT GGA GTT GTG GCA AAG CTG TCT CCC GCT AAC CAA AGG GTC CGC GCT GGG GTC ACC TGA GTG GGT CTC CCT CGG GAG GCC CTT TAG GTG CTG CCT CGG GCG GTG ACG AGT TTG ACA ATG TTC GTG GAA ATG TAG GCC TGC ACG GCT GGT CCT ATC TTG TCC CGG ACG CTG CCT CCT CCC GCC TCT GCA GGA TAT AGA CTG CAC CTT TTG CCC GGC GAG AGT TAG GAC GGA AAA GGC GAA GGC GGC GGC GGC TGT TGG GTG TCT CCC AGC ACT TGC GCC TGT ATA TGT GTA TCA CTG TGC GTG TAC GTA TGA AGC CTG CCA GCT GGA CCT GCC CGG TGG CAC CCC GGC TCT GAT CCT GAC CCA TGT CCT CCA CCG GGC GAA GTG GTA ACC TCC CTT CAG CTG CCA ACT GGC GTT TGT CCA GGT TGT GGC CCA CAT GTG CAA GCA ATT CAA CAG GTG AAC TGA GGC GAA CCC GTG GAG TGG AGG GGA TGC GGC CAG GGA GAA ATG TAA AGG AAA GAA TTG GGG CCC ATC CAT CCA CTG CAT AGC TGC CTG GCC CAG GAC AAA GCT GAT GTG CAG AGC ATC ATT GGA CTG CAG CGC TTC TTC GAG ACT AGA ATG AAT GAG GCT TTC GGT GAC ACG AAG TTC TCT GCC GTG CTG GTG GAG CCG CCC CCT ATG TCT TTG CCA GGA GCG GGA CTT TCA AGT CAG GAA CTC TCC GGC GGT CCT GGG GAT GGC CCA TAA GGC TGG AGT CCA CAT GGC CAG CCG TCA CTC GCC CTT TTC TCT GTT CTG AGC CCA CAT CCC CAC AGC CCC GGC GGC TTG ACG CCT ACG CCT TGG TGG CCA CAT CCT CCT GTT CCT CAC GAC ATG GTT TTC ACT TTT

5. His Tag: CATCACCATCACCATCATCAC
6. Stop Codon: TGA
7. SV40 Poly-A Signal:

TAAGATACATTGATGAGTTTGGACAAACCACAACTAGAATGCAGTGAAAAAAATGCTTTATTTGTGAAATTTGTGATGCTATTGCTTTATTTGTAACCATTATAAGCTGCAATAAACAAGTT

• Linear map visualization:

I exported the cassette as a FASTA file and uploaded it to Twist.

When deciding between Clonal Genes and Gene Fragments, the determining factors were: my cassette’s small size (even accounting for future sensor addition) and the practicality of working with a plasmid vector. For the backbone, I selected pTwist CMV OriP, which ensures high-level transient expression in mammalian cells, provides ampicillin resistance, and facilitates elevated protein expression.

Upon uploading, the platform flagged multiple repetitive sequence regions—consistent with the “high complexity” warning from the optimization tool. TRIM28 is an enzyme with functional domains; repetitive regions are expected. However, I’m concerned that my edits to bypass synthesis constraints may have compromised these functional domains.

Anyway, I downloaded the resulting GenBank file and re-imported it into Benchling.

• The final plasmid/vector is:

Part 5: DNA Read/Write/Edit

• 5.1 DNA Read: (i) What DNA would you want to sequence and why? In my proposed lupus autoregulatory circuit project, I would sequence cell-free DNA (cfDNA) from patient plasma to detect interferon-stimulated gene (ISG) signatures that predict disease flares. cfDNA sequencing offers a minimally invasive way to aquire information aboud patients’ in-real-time situation. Elevated IFN-α family actitvity would trigger preemptive TRIM28 circuit activation before clinical symptoms manifest—transforming lupus management from reactive to predictive. (ii) Sequencing technology choice In my expereience Ecuador, Illumina NovaSeq is the to go technology for biomedical research (second-generation sequencing). It is second-generation sequencing (sequencing-by-synthesis with reversible terminators), and works fine when needing to balances accuracy (Q30+ scores) with scalability—critical (e.g. when screening hundreds of rural patients). For the input (taken from the “QIAseq cfDNA All-in-One Kit Handbook”): Extract 10 ng cfDNA from 1 mL plasma (using silica-column kits compatible with low-input Twist workflows) Fragment to 300 bp via enzymatic shearing (avoiding sonication that degrades already-short cfDNA) Ligate Twist Universal Adapters with unique dual indexes (UDIs) to enable multiplexing of 96 samples per lane PCR amplify with 8 cycles (minimizing chimera formation noted in LeProust’s slides, p. 38) Base calling mechanism: Fluorescently labeled nucleotides (A/C/G/T with distinct dyes) are incorporated sequentially. After each cycle, lasers excite the dyes and cameras capture emission wavelengths. Software converts fluorescence patterns into base calls (A=green, C=blue, G=yellow, T=red), with quality scores reflecting signal-to-noise ratios. Output: 150 bp paired-end reads (300 bp total coverage per fragment) in FASTQ format. 5.2 DNA Write (i) What DNA would you want to synthesize and why? I would synthesize the TRIM28 expression cassette I designed in Part 3—a 1.2 kb sequence containing CMV promoter, Kozak sequence, human-optimized TRIM28 coding sequence, His tag, and SV40 poly-A signal. This cassette represents the effector module of my lupus circuit: when IFN-α levels rise during pre-flare states, TRIM28 would be expressed to SUMOylate and deactivate IRF-7 (Liang et al., 2011), breaking the pathogenic IFN-α amplification loop without broad immunosuppression. Unlike current therapies (e.g., anifrolumab antibody injections requiring monthly clinic visits), this circuit could provide continuous, autonomous regulation—critical for patients in Chimborazo province who travel 6+ hours for care. (ii) Synthesis technology choice I would use Twist Bioscience’s silicon-based phosphoramidite synthesis.

1. In silico design: Finalize cassette in Benchling with codon optimization for human cells (avoiding rare codons like AGG that stall ribosomes)
2. Chip synthesis: 150-mer oligos synthesized in parallel on silicon wafer (1 million oligos/chip)
3. Enzymatic assembly: i would opt to join independtly synthesized oligos into a full-length cassette.
4. Error correction: there must be a MutS protein to bind mismatched bases for removal
5. As stablished, I choosed cloning as to make it easier for the vector (pTwist CMV OriP) to reach mammalian (host cells) expression. A limitation encountered was that during my simulation (Part 4), Twist flagged my TRIM28 sequence as “high complexity” due to repetitive domains. This reflects a fundamental constraint: phosphoramidite chemistry has ~99% coupling efficiency per cycle (LeProust slides), making synthesis of long repeats error-prone. For my final project, I’d need to: -Break repeats with synonymous codon substitutions (preserving amino acid sequence) -Use Twist’s Ultra-complex Genes service (which is costly!!) for sequences with homopolymers >20 bp Accept that some functional domains may require post-synthesis assembly via Golden Gate cloning 5.3 DNA Edit (i) What DNA would you want to edit and why? My project is not about directly editing the patients DNA, as I am very concerned about incidentally damaging the patient. What I aim to do is to provide an external aid that is not as invasive as the current pharmacological guidelines. If I would given to choose what DNA to directly edit, I would be interested in modifying plants to become natural biorreactors: instead of producing their wildtype fruits (e.g. compounds required for the development of the seed), they could be edited to produce an specific compound (such as insulin for example).

References Baker T, Sharifian H, Newcombe PJ, et al. Type I interferon blockade with anifrolumab in patients with systemic lupus erythematosus modulates key immunopathological pathways in a gene expression and proteomic analysis of two phase 3 trials. Annals of the Rheumatic Diseases. 2024;83:1018-1027. Liang Q, Qi H, Hou F, et al. TRIM28 is a SUMO E3 ligase for IRF7. Journal of Biological Chemistry. 2011;286(37):32448-32456. McCarty NS, Graham AE, Studena L, et al. Rapid and scalable characterization of CRISPR technologies using an E. coli cell-free transcription-translation system. Molecular Cell. 2023;83(1):1-15.

Week 3— Lab Automation

Python Script for Opentrons Artwork

After making a copy of the collab doc into my drive, I searched for a reference image for my design. In the Designer Cells node there are two color options: red and green. So, I looked for a minimalistic design that could look good in green, my favorite color. I ended up choosing a hummingbird, as they come oftentimes to my garden to feed on the flowers’ nectar.
At the Automation Art Interface, there is the function of submitting an online image directly coping its direction, and it is suggested that its backdrop is white. The following is the selected reference image that meets the previous criteria:

(This is the image’s web direction/link pasted in the Automation Art Interface)

With it i obtained this incredibly beautiful and complex design:

I tinkered manually with the dots and with the default settings (Brightness, Contrast and Saturation) to modify the original suggested design, plus I changed all the dots to the “mClover3” color as I initially intended. After that, the coordinates I obtained were:

mclover3_points = [(23.1,27.5), (18.7,25.3), (20.9,25.3), (29.7,25.3), (16.5,23.1), (27.5,23.1), (-12.1,20.9), (-5.5,20.9), (12.1,20.9), (23.1,20.9), (-34.1,18.7), (-31.9,18.7), (-29.7,18.7), (-27.5,18.7), (-25.3,18.7), (-23.1,18.7), (-20.9,18.7), (-18.7,18.7), (-16.5,18.7), (-14.3,18.7), (9.9,18.7), (18.7,18.7), (20.9,18.7), (-3.3,16.5), (7.7,16.5), (25.3,16.5), (-16.5,14.3), (-14.3,14.3), (-1.1,14.3), (5.5,14.3), (14.3,14.3), (16.5,14.3), (25.3,14.3), (-12.1,12.1), (1.1,12.1), (12.1,12.1), (23.1,12.1), (1.1,9.9), (14.3,9.9), (20.9,9.9), (-1.1,7.7), (1.1,7.7), (20.9,7.7), (1.1,5.5), (18.7,5.5), (-9.9,1.1), (14.3,1.1), (-9.9,-1.1), (7.7,-1.1), (12.1,-1.1), (-7.7,-3.3), (9.9,-3.3), (-3.3,-7.7), (-1.1,-9.9), (12.1,-9.9), (14.3,-9.9), (14.3,-12.1), (18.7,-12.1), (9.9,-14.3), (18.7,-14.3), (7.7,-16.5), (14.3,-16.5), (20.9,-16.5), (7.7,-18.7), (20.9,-18.7), (7.7,-20.9), (16.5,-20.9), (20.9,-20.9), (9.9,-23.1), (18.7,-23.1), (20.9,-23.1), (14.3,-25.3), (18.7,-25.3), (16.5,-27.5), (16.5,-29.7)]

Returning to the Google collab document, scrolling to the last celd, below the Your code comment, I copy pasted the tuple containing the coordinates. Below the pasted tuple, on the same level of indentatio, I defined the well for retrieving the green color based on the posted examples. Then, to dispense the “colorant” into the precise coordinate it was needed, I started with the following lines of code:

pipette_20ul.pick_up_tip()    # to make the robot pick up a pipette tip to start dispensing
pipette_20ul.aspirate(len(mclover3_points) * 1.2, location_of_color('Green')) # to pick up enough volume of the “colorant” for all the coordinates’ dots.
for x, y in mclover3_points:
        adjusted_location = center_location.move(types.Point(x=x, y=y)) # to let the robot correctly identify the placement of the coordinates
        dispense_and_detach(pipette_20ul, 1, adjusted_location)  # to dispense 1µL per dot
pipette_20ul.drop_tip() # to drop the tip at the end of the task

I selected Run all and after the run, an error message popped up: I was calling pipette_20ul.pick_up_tip(), which didn’t exist. The run stopped there, and I accepted the help of the Gemini assistant to understand where the mistake lay:

Gemini observed that I had an indentation error when pasting the tuple with my coordinates, due which I had called for pipette_20ul.pick_up_tip() outside of def run(protocol). From that point on, I got the indentation for my code wrong. I corrected this by using tab at the start of every code line, and clicked on Run all again. However, for the second time an error message popped up:

Gemini summed up the mistake in that I had coded pipette_20ul.aspirate(len(mclover3_points) * 1.2, location_of_color('Green')) forgetting that the robot was using 20 ul tips. With the current line of code, I was taking 90 ul at the start of the task, which was physically impossible. Gemini suggested the next edit to correct the mistake while guaranteeing that the robot would not run out of “colorant”:

# Aspirate more liquid every 20 dispenses (or when starting) to avoid exceeding pipette capacity
for i, (x, y) in enumerate(mclover3_points):
  # Aspirate more liquid every 20 dispenses (or when starting) to avoid exceeding pipette capacity
  if i % 20 == 0:
    # Calculate the volume to aspirate: either 20uL (max for p20) or less if fewer than 20 points remain
    volume_to_aspirate = min(20, len(mclover3_points) - i)
    if volume_to_aspirate > 0: # Only aspirate if there's something to aspirate
      pipette_20ul.aspirate(volume_to_aspirate, green_well)

I hit the Run all option again and this time the code ran smoothly, and I could visualize my final design:

How AI aided me?: I am familiar with Python coding (but not with the Opentron API libraries). Gemini helped me to understand the logic/algorithm behind the coding. And this AI also became proofreader, noticing details that I missed but that directly affected my run.

Post-Lab

1. Finding a published paper that utilizes Opentrons or an automation tool to achieve novel biological applications:

I found a published review paper: Automation and artificial intelligence in filamentous fungi-based bioprocesses: A review, that showcases how continuous data acquisition through intelligent sensors combined with Machine Learning (A.I) and automatization tools and protocols could boost fungi-based bioprocesses.

Taken from the paper by (Wainaina & Taherzadeh, 2023)

On their subchapter 3.4, they present the concept of the smart bioreactor in the context of filamentous fungi. The authors propose that such bioreactors should integrate the existing infrastructure of a bioreactor with the concept of a hierarchical structure control system (HSCS). HSCS is a “set of devices and software organized in a tiered manner” where “the layers on the higher levels are designed to perceive theoretical models and execute planning while the lower ones perform local tasks…”(Wainaina & Taherzadeh, 2023). The authors explain that automated sensors could help detect fermentation rate and related parameters, so researchers could monitor the bioreactor progress and make adjustments in real-time. This grabbed my attention, being the first time that I associated automation with bioreactors. HSCS prototyping could be a door to upscaling bioreactor production on low funded startups.

2. Intended automation tools usage in my final project:

During Week 2, when I designed my TRIM28 expression cassette in Twist, the platform flagged my sequence as “high complexity” due to repetitive sequence parts (which I presume are part of the RBCC region’s domains). I have heard before that this is a common challenge when working with natural enzyme sequences that contain tandem repeats. This technical barrier directly impacts my goal of developing an autoregulatory circuit for lupus patients with a poor initial prognosis. I had to manually delete several high repetition regions for Twist being able to accept my order, and I wondered how this change could alter the original functionality of the enzyme. I also started to question the opposite: how much I could simplify the sequence and have the resulting protein performing its original function. A further simplified sequence would be cheaper and faster to sequence, optimizing at least one step of the process needed to reach my project´s goal.

So, I thought that I can use Opentron to make a trial in this regard:

I could order 3 to -5 variants of my TRIM28 construct, each with more high repetition sites manually deleted. However, to each construct I would add a ubiquitination sensor (Qin et al., 2022; Choi et al., 2019).

In the presence of TRIM28 and IRF7 ubiquitination should occur as part of IRF7 degradation process. I could gauge which construct is working by measuring the ubiquitination rate. Ubiquitination occurs naturally in the cells, so a control group of cells (without the construct expressing continuously TRIM28) should be added to the well, to have a contrast that can showcase the “extra” ubiquitination activity in the transfected cells. Literature suggests that after a cell culture is transfected, analysis should start after 24- 48 hours (only then cells could be harvested) (Qin et al., 2022).

I would use the Opentron robot to automate the process of adding each different construct to a well in a well plate containing mammalian cell lines. After robot preparation, cells would be manually transferred to a standard CO₂ incubator — a hybrid workflow that reduces human error in small-volume transfers. After the incubation period, the well plate would return to the Opentron robot to trypsinize cells for 2 purposes: 1) make a subculture; 2) Perform a fluorescence test.

1. The cell subcultures of each cell group would be used to perform a western blot to determine the presence and relative abundance of TRIM28. The Opentron website mentions that the western blotting process could be automated , so this part could be left to the robot to increase throughput.
2. There is also a link in the Opentron website about Fluorescence-Activated Cell Sorting (FACS) Sample Prep , however, in this protocol the robot adds staining fluorescent antibodies, which is not needed or wanted for my current design. If this step could be omitted, and some other steps edited to enhance coherence within steps, then the fluorescence assay could be automated too. If not, then after cell trypsinization the cell culture remaining in the well plate could be manually taken to the fluorescence microscopy equipment.

References:

• Choi, YS., Bollinger, S.A., Prada, L.F. et al. High-affinity free ubiquitin sensors for quantifying ubiquitin homeostasis and deubiquitination. Nat Methods 16, 771–777 (2019). https://doi.org/10.1038/s41592-019-0469-9
• Qin, W., Steinek, C., Kolobynina, K., Forné, I., Imhof, A., Cardoso, M. C., & Leonhardt, H. (2022). Probing protein ubiquitination in live cells. Nucleic acids research, 50(21), e125. https://doi.org/10.1093/nar/gkac805
• Wainaina, S., & Taherzadeh, M. J. (2023). Automation and artificial intelligence in filamentous fungi-based bioprocesses: A review. Bioresource Technology, 369(128421), 128421. https://doi.org/10.1016/j.biortech.2022.128421