Week 1 HW: Principles and Practices

1) Biological engineering application / tool

One biological engineering tool I would like to develop is a programmable biological computing platform based on synthetic genetic circuits that can sense biological inputs, perform simple computational operations, and produce interpretable outputs such as a visible color change or measurable signal. This idea comes from my interest in using synthetic biology as a medium for computation that could complement or eventually reduce reliance on energy-intensive electronic systems.

Rather than using biological systems only as sensors, I am interested in engineering circuits that can process multiple inputs and perform basic forms of computation within living or cell-free systems. For example, a biological circuit could detect a combination of biological markers and compute a response based on their relationships, producing an output that reflects the processed information rather than a single input. These types of systems could act as early models for more advanced biological computation and could eventually contribute to sustainable technologies that merge biology and information processing.

I am also interested in how neural network concepts and quantum computing approaches might inform the design of such biological circuits. Neural networks provide a framework for pattern recognition and adaptive systems, while quantum and quantum-inspired optimization methods may help solve complex design problems involved in assembling reliable biological circuits. Although these approaches are still emerging, integrating them with synthetic biology could help advance a future in which biological systems play a role in computation, diagnostics, and personalized medicine.

In the long term, platforms like this could support more personalized and environmentally sustainable medical technologies by allowing biological systems to process information locally and respond early to complex biological states. However, because tools that combine sensing, computation, and biological data processing could be misused or misinterpreted, it is important to consider governance and ethical safeguards alongside technical development.

2. Governance and policy goals for an ethical future

Main goal

Enable beneficial innovation in biological computing while preventing unsafe deployment, privacy violations, and misleading medical use.

Sub-goal 1: Reduce dual-use and misuse risk (non-malfeasance)

Because programmable biocircuits and design workflows can lower barriers to engineering biology, governance should reduce the chance they’re repurposed for harm (intentional misuse) and reduce accidental risks (lab accidents, unintended releases). This is the core concern in synthetic genomics governance options and WHO dual-use governance.

Sub-goal 2: “Trustworthy outputs” for health contexts (avoid harm from errors)

If biocircuits are ever used to inform health decisions (even as screening tools), governance should prevent harm from false positives/negatives, unclear limitations, and overconfident marketing. The Bioethics Commission stresses ongoing risk analysis, transparency, and public trust; this fits that logic.

Sub-goal 3: Protect autonomy, privacy, and consent

Biological systems that sense/compute on biological states could enable non-consensual testing or sensitive inference. Governance should promote consent norms, privacy-preserving design, and clear boundaries for acceptable uses—consistent with ethics frameworks emphasizing transparency, stewardship, and justice/fairness.

Sub-goal 4: Promote equitable access and responsible innovation

Ensure benefits (diagnostics, sustainability, safer computing) are broadly shared and not limited to wealthy settings; align incentives so safety and equity are not afterthoughts. This aligns with “societal goals” framing in the White House biotech R&D goals report.

3. Governance actions

Action 1 — “Architecture/code” + industry norm: sequence screening + provenance for risky designs

Actors: DNA synthesis & gene/genome services, protein design platforms, funders, and journals.

Purpose: Today, many safeguards rely on institutional biosafety review and some DNA-order screening, but AI/protein design and synthesis are moving fast. A proposed change is to strengthen screening and add provenance/traceability for high-risk designed sequences so dangerous designs are harder to order and easier to flag. This mirrors the biosecurity direction argued in the Baker & Church Science editorial and fits WHO-style biorisk mitigation.

Design:

1. Require/expand sequence screening for orders (DNA and other relevant synthesis inputs) using updated threat databases and better detection of modified/obfuscated sequences.

2. Store minimal audit logs (provenance) for flagged/high-risk categories with clear governance on who can access logs and under what conditions.

3. Tie compliance to funding/journal policies (e.g., “must use screened providers”) and to procurement requirements. (This is similar in spirit to governance options discussed for providers/users in synthetic genomics governance thinking.)

Assumptions:

1. Assumes screening algorithms keep up with rapid design methods.

2. Assumes there’s agreement on what counts as “high-risk” and how to avoid overblocking legitimate research.

3. Assumes audit logging can be done without creating new privacy/security problems.

Risks of failure & “success”:

Failure: adversaries route around regulated providers; screening misses novel threats; smaller labs get locked out.

Success: more surveillance infrastructure than necessary; chilling effect on open science; inequity if only wealthy groups can access compliant supply chains.

Action 2 — “Norms + law” in research & community biology: tiered biosafety/biosecurity readiness + external support

Actor: universities, community labs (DIYbio spaces), iGEM-like education programs, local regulators, biosafety professionals.

Purpose: Right now, safety practices vary widely across institutions and community spaces. The change is a tiered, “right-sized” readiness system: clear requirements for training, risk assessment, and project review that scale with risk—especially for projects involving novel circuits, pathogen-adjacent work, or anything that could be misused. This aligns with the Bioethics Commission’s call for ethics education and governance that anticipates DIY contexts, and with WHO’s emphasis on governance tools/mechanisms for biorisk and dual-use.

Design:

1. Adopt a tiered checklist + review model (low-risk fast path; higher-risk requires review).
2. Require documented training aligned with recognized biosafety references and iGEM-style responsibility practices (risk forms, safety planning, escalation paths).
3. Build an “ask-an-expert” escalation channel for community labs (modeled on DIYbio’s biosafety expert portal concept).
4. Encourage periodic external audits for higher tiers (WHO mentions external audit as a mechanism).

Assumptions:

1. Assumes community labs will opt in if requirements aren’t too burdensome.
2. Assumes there are enough trained biosafety professionals to support audits/advice.
3. Assumes tiers map well to risk in fast-changing tech areas.

Risks of failure & “success”:

Failure: becomes box-checking; pushes experimentation underground; inconsistent enforcement.

Success risk: could reduce accessibility for low-resource groups unless paired with funding/support; may centralize power in institutions and exclude community innovation.

Action 3 — “Market + regulation” for health-facing use: pre-market validation + truth-in-marketing + post-market monitoring

Actor: companies (including startups), healthcare regulators, clinical partners, insurers, app platforms.

Purpose: Today, the jump from “cool demo” to “health claim” is where harm happens. For biocircuits that might be used in personalized medicine pathways, require that products can’t be marketed as diagnostic/therapeutic guidance without evidence standards and clear labeling of limitations. This supports the Bioethics Commission’s emphasis on transparency, ongoing risk analysis, and public trust.

Design:

1. A staged pathway: research-use-only → screening/decision-support → clinical diagnostic, with increasing evidence requirements.
2. Clear performance metrics (sensitivity/specificity, failure modes, bias across populations) and user-facing communication.
3. Post-market monitoring for real-world failures and misuse (aligns with “ongoing” oversight logic rather than one-time approval).
4. Align incentives: procurement and reimbursement favor validated tools (a “market lever”), consistent with “biotech to further societal goals” framing.

Assumptions:

1. Assumes regulators can classify and evaluate novel “bio-computing” tools cleanly.
2. Assumes companies won’t avoid regulation by making vague “wellness” claims.
3. Assumes validation datasets represent diverse populations.

Risks of failure & “success”:

Failure: regulatory gaps lead to misleading products; over-regulation slows beneficial innovation.

Success risk: big companies with money for trials dominate; smaller innovators struggle unless there are grants/partnership pathways.

4. Scoring governance actions

5. Reflection on trade-offs

Based on the scoring, I would prioritize a combination of Option 2 and Option 1, while treating Option 3 as a later-stage priority once the technology begins to move toward health-facing applications. Option 2, which proposes a tiered biosafety and responsibility framework for academic and community laboratories, stands out as the most feasible and immediately impactful starting point. It scored highest for fostering lab safety, improving response capacity, and not impeding research, because similar training and review systems already exist in many research environments and can be expanded without excessive cost. Strengthening education, risk assessment, and oversight in a tiered way allows governance to scale with the level of risk rather than applying overly strict requirements to all projects. This aligns with governance approaches that emphasize responsible stewardship, ethics training, and adaptive oversight while still supporting innovation. Option 1, which involves sequence screening and provenance tracking for higher-risk designs, complements this by addressing upstream biosecurity risks that cannot be fully mitigated through local lab training alone. Screening and traceability can help prevent misuse and enable investigation if something goes wrong, making it an important second priority that strengthens prevention at the supply-chain level.

Option 3, focused on pre-market validation and clear rules for health-related claims, becomes essential once biological computing tools begin to be used in medical or personalized medicine contexts. Although it scored lower on feasibility and cost because regulatory validation can be burdensome and slow, it is critical for preventing harm from inaccurate or misleading outputs and for maintaining public trust. For that reason, I would treat Option 3 as a second-phase priority: not the first governance action for early-stage research, but something that should be planned early so that clear evidence standards exist before technologies are deployed in real health settings.

Several trade-offs influenced this prioritization. Strong screening and regulatory measures can improve safety and security but may also increase costs, slow research, or concentrate power in larger institutions that can more easily comply with requirements. Conversely, focusing only on education and local oversight could leave gaps in system-wide protection if risky materials can still be obtained without screening. The combined approach of Option 2 and Option 1 balances these concerns by strengthening safety culture and responsibility at the research level while also adding upstream safeguards to reduce misuse. However, this approach assumes that risk categories can be defined clearly, that screening systems will keep pace with advances in design technologies, and that enough biosafety expertise exists to support tiered review processes. There is also uncertainty about how regulators will classify and evaluate emerging biological computing tools that blur boundaries between research, diagnostics, and wellness technologies.

My recommendation is directed primarily toward university leadership, biosafety offices, research funders, and community biology networks, because these actors are well positioned to implement tiered safety frameworks and require the use of screened synthesis providers. A secondary audience includes biotechnology companies and synthesis provider consortia, whose cooperation would make screening and traceability more consistent across the field. Finally, as the technology matures toward personalized medicine applications, health regulators and clinical partners would become key audiences for implementing validation and marketing standards. Overall, prioritizing Option 2 and Option 1 provides a practical and balanced foundation for responsible development, while Option 3 ensures that future clinical uses are introduced in a safe and trustworthy way.

WEEK 2 LECTURE PREP

Homework Questions from Professor Jacobson

QUESTION 1

From the slide comparing chemical vs biological DNA synthesis, the error-correcting polymerase is shown with an error rate ~ 1 in 10⁶ (10⁻⁶) per base. The slides list the human genome length as ~3.2 Gbp. If you multiply that out: 3.2×10⁹ bases × 10⁻⁶ ≈ 3.2×10³, so ~3,200 errors per genome copy at that raw polymerase error rate. That’s a big discrepancy. Biology reduces discrepancy in layers, including:

1. Proofreading exonuclease activity (noted on the polymerase slide as proofreading/error-correcting exonuclease functions).

2. Post-replication mismatch repair, e.g., the slide explicitly mentions the MutS repair system as an “Error Correction” mechanism.

3. Plus broader cellular responses (conceptually): damage repair pathways, cell-cycle checkpoints, and removal of heavily damaged cells.

QUESTION 2

The slides give an average human protein coding length of around 1036 bp. Because the genetic code is redundant since most amino acids have multiple codons, the number of distinct DNA sequences that could encode the same protein is very large. In practice, many of these sequences don’t work well because the same protein does not equal the same expression.

Some reasons why not all codes work in practice are:

1. The tRNA abundance differs by organism which therefore affects translation speed and accuracy.
2. GC content and repeats which can make DNA hard to synthesize or clone. It can cause recombination instability.
3. mRNA secondary structure can block ribosome binding or slow translation.
4. Restriction sites or assembly constraints.

Homework Questions from Dr. LeProust

QUESTION 1

The standard workhorse method is solid-phase phosphoramidite chemical DNA synthesis, shown as the “Phosphoramidite DNA Synthesis Cycle” in the slides.

QUESTION 2

Because direct chemical synthesis is stepwise: each added base has <100% coupling efficiency, so the yield of full-length product drops exponentially with length. It also accumulates deletions or substitutions, and longer products are harder to purify away from the huge mixture of truncated failure products.

QUESTION 3

The compounded stepwise yield becomes extremely low, and the error burden becomes too high to get enough correct full-length molecules. Instead, many shorter oligos can be synthesized, then assembled into longer DNA using Gibson Assembly, and later the sequence gets verified and corrected if needed.

Homework Question from George Church

The 10 essential amino acids for animals are:

1. Arginine
2. Histidine
3. Isoleucine
4. Leucine
5. Lysine
6. Methionine
7. Phenylalanine
8. Threonine
9. Tryptophan
10. Valine

(Google search)

Lysine contingency is a scientific procedure created for Jurassic Park, which consists of using lysine dependence and availability as a safety control. The fact that lysine is essential for animals doesn’t automatically make it a strong containment strategy. Animals need lysine from diet, but lysine is still common in many real-world environments because it’s present in food, biomass, and is produced by microbes/plants. So lysine dependence alone could be a weak containment barrier outside carefully controlled settings. This makes me think lysine-based containment could work as one layer in a broader safety plan, but it likely shouldn’t be the only line of defense.