Week 1 HW: Principles and Practices

First, describe a biological engineering application or tool you want to develop and why.

I want to develop a closed loop pipeline for peptide engineering that uses Feynman–Kac steering to control diffusion-based protein generation at inference time. The goal is to go beyond zero-shot prediction and instead build an automated engineering cycle that repeatedly:

• proposes peptide/mini-protein candidates,
• captures experimental readout (binding, activity, stability, etc.),
• turns those measurements into reward signals,
• uses FK steering to bias the next round of generative sampling toward better candidates without needing to retrain the underlying diffusion model

This is inspired by FK-steering approach which wraps a diffusion protein generator with a sampling scheme so trajectories are continuously reweighted toward user-defined rewards, which in this case, is the experimental readout.

Peptides are a good choice for this project as they are often fast to synthesize and test, making them compatible with iterative lab loops. However, many properties of peptides we care about (solubility, stability, expression, off-target behavior) can be hard to optimize from prediction alone so a wet-lab loop is attractive. Functionally, they can serve as binders, inhibitors, diagnostic reagents, or modular parts in synthetic biology pipelines.

As a concrete MVP within this class, I hope to learn how to perform the wet lab experiments associated to this project and finish at least 1 cycle. In the medium term, I would like to run comparisons between different computational approaches like simple finetuning or RL. In the long term, I would like to utilizie this method to discover therapeutic proteins.

Next, describe one or more governance/policy goals related to ensuring that this application or tool contributes to an “ethical” future, like ensuring non-malfeasance (preventing harm). Break big goals down into two or more specific sub-goals.

Closed loop design could be repurposed to create harmful biomolecules. Governance should reduce the probability of both deliberate misuse and accidental creation of dangerous function. Thus, one major goal would be to prevent misuse. As sub goals, the following may be good options:

• Ensure the system does not optimize toward harmful or restricted targets/functions.

• Reduce the chance that hazardous sequences are synthesized without review.

• Ensure that there are audit trails and responsible-use norms.

Next, describe at least three different potential governance “actions” by considering the four aspects below (Purpose, Design, Assumptions, Risks of Failure & “Success”).

I propose three governance actions spanning institutional review, synthesis controls, and a logging infrastructure.

Option 1: Institutional Review

• Purpose: Add structured risk assessment before synthesis, target changes, or new reward functions in academic protein design projects.
• Design: One-page checklist covering target protein class, reward function, synthesis plan, and screening. Projects triggering high-risk criteria (regulated agents, virus optimization) require formal oversight.
• Assumptions: Small review gates and enforce good record keeping practices
• Risks: Could push students to under-report. If too strict, it may slow down R&D>

Option 2: Synthesis Controls

• Purpose: Require synthesis vendors to use functional or homology-based screening.
• Design: Institutions only purchase from vendors who screen orders and verify customers
• Assumptions: It is possible to do screening meaningfully well to reduce risk
• Risks: The screening needs to be highly accurate to catch edge cases which could have massive negative effects

Option 3: Logging Infrastructure

• Purpose: Create a secure shared database that tracks when AI tools generate protein designs
• Design: Logging of AI tools and cross-referencing of orders.
• Assumptions: Confidentiality and transparency is balanced
• Risks: Security or confidentiality concerns from hacking or from sensitive IP

Last, drawing upon this scoring, describe which governance option, or combination of options, you would prioritize, and why. Outline any trade-offs you considered as well as assumptions and uncertainties.

In order of priority:

• Option 1: This option can arguably be implemented the fastest. MIT already has the safety infrastructure (IBC, EHS) to build on. As a leading institution in AI protein design, MIT can set standards that others follow. A well-designed, lightweight review process could become a widely adopted model.
• Option 2: The existing government framework provides a strong template with vendor screening, customer verification, and reporting requirements. However, this depends on federal action and industry cooperation beyond MIT’s control. MIT can help by researching better screening algorithms and influencing governement gold standards.
• Option 3: If this project becomes a widely used system, tracking who designed what becomes relatively easy. However, the system will have to be designed extremely well to be scalable, secure, transperent yet confidential.

Tradeoffs:

• Speed vs. safety
• Open science vs. closed science
• Transparent vs. confidential

Key Uncertainties:

• How manageable it is to manually gate research directions.
• How well screening actually works against deliberate misuse.
• How feasible it is to design a logging system everyone is happy with.

Reflecting on what you learned and did in class this week, outline any ethical concerns that arose, especially any that were new to you. Then propose any governance actions you think might be appropriate to address those issues. This should be included on your class page for this week.

Unfortunately, I was ill this week so I was not able to attend class.