Week 1 HW: Principles and Practices

Homework Week 1: Class Assignment

Biological Engineering Application

First Steps towards “Intelligence in a (warehouse)-dish”

Guided by the vision of building a biological general computing system, the goal of the proposed tool is to provide a minimal, yet replicable brain organoid based system, that can be engineered to exhibit controllable, learning-like signal processing behaviour. The system consists of 3 conceptual parts (input - computation - output), that manifest in 2 integrated physical devices.

Firstly, there is the 3D organoid culture chamber that handles the computation. Brain organoids, based on iPSC cultures, with a diameter below 500 micrometer, containing less than 100,000 cells can be manufactured with a high degree of standardization and scalability. More recent research allowed the long-term culture of brain organoids exceeding one year, displaying spontaneous electrophysical (re)activity, show extensive myelination, and can be enriched with several relevant cell types, e.g. oligodendrocytes, microglia, and astrocytes [1, 2, 3, 4].

Second is the input-output system that handles the input and output functions. As current research focuses on avascular brain organoids that require delivery of nutrients via diffusion, a higher order, more complex brain organoid requires brain vasculature. The microfluidic system mimics above-mentioned vasculature and contributes to the development of higher-order brain organoids. Additionally, the system can deliver and record chemical signals in a spatiotemporal manner [1, 5, 6]. Another activation mechanism lies in 3D microelectronic arrays. These allow precise stimulation and recording of spatiotemporal signals across the entire surface of the brain organoid [7].

Therefore, combining advances in 3D culture of brain organoids, with a combination of microfluidics and microelectronic arrays, poses an exciting research avenue and aims to contribute to the research topic of Organoid Intelligence (OI) [1].

Citations

1. Smirnova, L., Caffo, B.S., Gracias, D.H., Huang, Q., Morales Pantoja, I.E., Tang, B., Zack, D.J., Berlinicke, C.A., Boyd, J.L., Harris, T.D. and Johnson, E.C., 2023. Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Frontiers in Science, 1, p.1017235
2. Pamies, D., Barrera, P., Block, K., Makri, G., Kumar, A., Wiersma, D., Smirnova, L., Zhang, C., Bressler, J., Christian, K.M. and Harris, G., 2016. A human brain microphysiological system derived from induced pluripotent stem cells to study neurological diseases and toxicity. Altex, 34(3), p.362.
3. Monzel, A.S., Smits, L.M., Hemmer, K., Hachi, S., Moreno, E.L., van Wuellen, T., Jarazo, J., Walter, J., Brüggemann, I., Boussaad, I. and Berger, E., 2017. Derivation of human midbrain-specific organoids from neuroepithelial stem cells. Stem cell reports, 8(5), pp.1144-1154.
4. Marton, R.M., Miura, Y., Sloan, S.A., Li, Q., Revah, O., Levy, R.J., Huguenard, J.R. and Pașca, S.P., 2019. Differentiation and maturation of oligodendrocytes in human three-dimensional neural cultures. Nature neuroscience, 22(3), pp.484-491.
5. Grigoryan, B., Paulsen, S.J., Corbett, D.C., Sazer, D.W., Fortin, C.L., Zaita, A.J., Greenfield, P.T., Calafat, N.J., Gounley, J.P., Ta, A.H. and Johansson, F., 2019. Multivascular networks and functional intravascular topologies within biocompatible hydrogels. Science, 364(6439), pp.458-464.
6. Sabatini, B.L. and Tian, L., 2020. Imaging neurotransmitter and neuromodulator dynamics in vivo with genetically encoded indicators. Neuron, 108(1), pp.17-32.
7. Huang, Q., Tang, B., Romero, J.C., Yang, Y., Elsayed, S.K., Pahapale, G., Lee, T.J., Morales Pantoja, I.E., Han, F., Berlinicke, C. and Xiang, T., 2022. Shell microelectrode arrays (MEAs) for brain organoids. Science advances, 8(33), p.eabq5031.

Governance Goals

Harm reduction

Minimize the harm on the biological system through careful research design, consideration of biological responses and sensibility at the intersection of the research goal and societal norms.

Biological harm reduction emphasizes preserving the physiological integrity of the organoid system on the biological level. The intent is to avoid inducing unnecessary stress, damage or pathological stress in living tissue and to ensure that experimental interactions remain compatible with healthy biological function.

Limitation of harm arising from emergent properties seeks to prevent unintended transitions toward higher-order dynamics that could raise ethical concerns, as organoids exhibit more complex, self-organizing behavior. This includes constraining system complexity and maintaining clear boundaries on the duration and scope of experiments, as well as ensuring that organoids are not maintained beyond their justified research purpose.

Donor rights

Ensuring that individuals that contribute their biological samples for research retain fair rights, autonomy and benefits and are protected again extractive behaviour of third parties though the entire lifecycle and downstream applications.

Transparent disclosure of organoid intelligence research ensures that donors clearly understand that their biological material may be used to generate brain organoids for learning-like signal processing and computational experimentation.

Fairness in benefit sharing and protection against discrimination aims to include donors into the benefits generated by the research, nor expose them to harm through stigmatization, profiling, or inequitable use of data derived from their biological contributions. Donor participation should not create asymmetries where value is extracted without corresponding ethical consideration.

Privacy preservation protects donors from identity linkage, misuse or inappropriate inference of personal traits.

Access

Ensuring that access to the systems themselves as well as the associated knowledge and benefits are disseminated in a fair way, while extractive use attempts are actively prevented.

Equitable research access seeks to avoid concentration of organoid intelligence capabilities within a small number of research groups and institutions. The intent is to enable participation by a diverse range of (also non-scientific) communities.

Non-exclusive access tries to ensure that foundational biological systems and insights won’t be locked behind proprietary structures. The goal is to preserve openness at the level of core knowledge and enabling technologies.

Limiting extractive use ensures that access to sensitive biological data does not enable exploitation of contributing or downstream affected individuals. This goal emphasizes that organoid intelligence research should generate value that is aligned with societal benefit.

Governance Actions

Action 1: Technically Enforced Graduated Freedom

Purpose: Redefining the locus of governance from external to internal. The goal of this action is to embed harm reduction directly into the technical architecture, while still preserving scientific flexibility. Instead of imposing rigid limits, the system provides ethical “factory settings” that enable safe and broadly acceptable use by default, while allowing controlled exploration beyond these settings when justified.

Design: The organoid computing platform is developed with a set of default operating parameters, e.g. size, culture duration, stimulation intensity, and learning persistence. These can be used without additional ethical review. At the same time, a clearly defined subset of parameters is designated as research-variable, allowing researchers to intentionally explore higher complexity, longer duration, or altered learning dynamics. Deviations beyond default settings require explicit justification and appropriate ethical oversight, but are technically supported rather than prohibited. The system logs when and how parameters are modified.

Assumptions: It is assumed that most researchers will operate within default settings unless there is a genuine scientific reason to deviate.. It also assumes that technical transparency (rather than hard locks) is an effective governance lever.

Risks of Failure & “Success”: The model fails if defaults are treated without care rather than minimum safeguards, or if parameter variation becomes routine without oversight. Successful use of the system could create a false sense of ethical safety. There is also a risk that logging is perceived as surveillance, which in turn would discourage experimentation.

Action 2: Reciprocal Donor Stewardship

Purpose: Current consent frameworks mostly are a one-time action, offering limited protection against extractive use. This action proposes a reciprocal donor stewardship model, in which the collecting institution acts as a fiduciary to protect the donor’s interests, but also to maintain a two-way informational relationship. Donors are recognized as long-term stakeholders.

Design: Universities and biobanks adopt stewardship responsibilities as a condition for ethical approval and public funding. Donors opt into a structured relationship that includes regular high-level updates on relevant scientific developments as well as personalized notifications when findings derived from their samples may have health relevance.

Assumptions: This model assumes donors want an ongoing relationship. Furthermore it assumes that institutions can responsibly manage communication. It also assumes that research findings can be meaningfully categorized into general scientific updates versus personally relevant information.

Risks of Failure & “Success”: The model may fail if institutions lack the willingness to maintain long-term engagement. There is also a risk that donors misinterpret research signals as medical diagnoses, causing anxiety or harm. Successful implementation could blur the boundary between research and clinical care.

Action 3: Simple Public-Interest Licensing

Purpose: Biological computing moves toward commercialization, this creates the option that foundational technologies become locked behind exclusive or opaque licensing arrangements. The goal of this action is to preserve the public-interest while enabling rapid and practical commercialization, ensuring that ethical constraints do not themselves become barriers to innovation.

Design: Universities and spin-outs adopt standardized, plain-language public-interest licenses. These licenses are intentionally short, unambiguous, and easy to interpret, defining only a small number of clearly prohibited applications, while leaving all other commercial uses unrestricted. Investors and companies opt in upfront, gaining predictability.

Assumptions: This approach assumes that ethical constraints can be expressed in a small number of clear, enforceable prohibitions. It also assumes that companies and investors value legal certainty and speed of commercialization enough to accept modest limits on exclusivity and application scope.

Risks of Failure & “Success”: The model fails if prohibited-use categories are defined too broadly or too minimally. Conversely, “success” could lead to widespread adoption, which may normalize these constraints.

Scoring Matrix

Recommendation to ethics boards

For research on organoid intelligence, ethics boards should prioritize governance mechanisms that operate at the point of experimental design and focus on setting default use behaviour. Based on the policy goals of harm reduction, donor rights, and access, I recommend ethics boards focus primarily on Action 1 (Technically Enforced Graduated Freedom) for the experiment design and Action 2 (Reciprocal Donor Stewardship) ensuring a modern relationship management. Action 3 (Simple Public-Interest Licensing) will become more relevant in the near future, therefore it should be considered down the road.

First, ethics boards should require technically enforced ethical defaults for organoid intelligence systems. Rather than relying on lengthy binary approval decisions, projects should be judged on justification of any intended deviations. Second, ethics boards should transition from one-time consent towards reciprocal stewardship plans. These plans should treat donors as long-term stakeholders engaging in two-way communication when findings may be personally relevant. This strengthens donor autonomy and public trust without conflating research with clinical care. While ethics boards do not manage IP, they should recommend investigators to include public interest licensing in their research lifecycle.

The risk that overly cautious governance discourages legitimate research. These uncertainties argue for graduated, revisitable oversight rather than rigid prohibitions.

Use of Generative AI

Generative AI was used as a drafting aid throughout the development of this homework assignment. Specifically, it supported the structuring and refinement of complex ideas at the intersection of organoid intelligence, and governance, including the logical separation and articulation of policy goals and governance actions. The AI was used to iteratively clarify language and explore alternative framings, while all substantive ideas, judgments, and final decisions were made by the author.

Week 2 Lecture Prep

Prof. Dr. Jacobson

Nature’s machinery for copying DNA is called polymerase.

What is the error rate of polymerase?

The error rate is 1:10E6 (see page 8, right side “biological systhesis”)

How does this compare to the length of the human genome?

The human genome has a length of 3.2 Gbp or 3.2 x 10E9 (see page 10 diagram)

How does biology deal with that discrepancy?

The human body engages in a “proofreading” process (page 8). The error correction process, called “MutS Repair System” is described in detail in Lamers et al. Nature 407:711 (2000) (page 14)

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?

The average human protein is 1036 bp, equating to 1036 / 3 = 345 amino acids. On average 3 codons encode an amino acid, therefore the number of number of different DNA sequences encoding the same 345 amino acids is 3 x 10E345.

Dr. LeProust

What’s the most commonly used method for oligo synthesis currently?

Phosphoramidite chemistry. Page 2 illustrates the 4 step process. See the historical development from page 9.

Why is it difficult to make oligos longer than 200nt via direct synthesis?

Not directly found in the slides. Searching further ressources says the fundamental issue is cumulative coupling efficiency loss. Each cycle has a coupling efficiency of ~99–99.5%. The overall yield is the stepwise efficiency raised to the power of the oligo length: for a 200-mer at 99.5% coupling efficiency, the yield of full-length product is roughly 37%.

Why can’t you make a 2000bp gene via direct oligo synthesis?

Same rational as answer above. Decay is exponential, therefore the efficiency drops to negligible levels.

Prof. Dr. George Church

Using Google & Prof. Church’s slide #4: What are the 10 essential amino acids in all animals and how does this affect your view of the “Lysine Contingency”?

Phenylalanine (F), Valine (V), Threonine (T), Tryptophan (W), Isoleucine (I), Methionine (M), Leucine (L), Lysine (K), Histidine (H), Arginine (R)

Opus 4.6 Prompt: How does this relate to Lysine Deficiency?

The connection to Jurassic Park wasn’t clear to me, as i haven’t seen the movie. Therefore here is Opus answer (which was an amusing read): “The “Lysine Contingency” from Jurassic Park

In the film/novel, the engineers claim dinosaurs were designed to be lysine-deficient — unable to produce lysine — so they’d die without supplemental lysine provided by the park, serving as a biocontainment mechanism.

Why It Doesn’t Work The “contingency” is essentially meaningless as a containment strategy because lysine is already an essential amino acid for all animals. No animal synthesizes its own lysine. The dinosaurs wouldn’t be any more dependent on external lysine than a cow, a bird, or a human. Any environment with protein-containing food sources (plants, prey, insects) would supply adequate lysine. The dinosaurs would simply eat and obtain lysine the same way every other animal does.

Connection to Church’s Work Interestingly, slide #4’s emphasis on “knowing your monomers” and the references to non-standard amino acids (NSAAs) and xenomicrobiology (genetic code engineering, referenced from the 2016 Microb. Biotech paper) point to what actual biocontainment might look like. As shown in slides 39–42, Church’s lab has built Genomically Recoded Organisms (GROs) that depend on NSAAs like BipA — amino acids that truly do not exist in nature. This is a real lysine contingency done right: organisms that require a synthetic amino acid unavailable in any natural food web, providing genuine genetic and metabolic isolation.

So Crichton’s fictional biocontainment was flawed at the basic biochemistry level, but Church’s lab has essentially built the version that actually works.”