One instruction.
Build the project.
Week 13 had no questions, no protocol, no Q&A. The assignment was: "Work on your Final Project. Present it May 13 (Committed Listeners)." This page documents what that work looked like — the deliverables produced, the design decisions made, and the feedback that made the project more rigorous at each iteration.
The week's lecture was Renee Wegrzyn (former ARPA-H Director) on AI, SynBio, and scaling health innovation. The recitation covered Golden Gate Assembly — a Type IIS restriction enzyme-based method directly relevant to the modular cassette architecture of the Füzi Poiesis plasmid designs, and to the R-M shielding strategy applied in Sub-Aim 1.1.
Four deliverables.
All computational. All documented.
The demonstration chassis is E. coli K-12 MG1655 — chosen because its kinetic parameters are fully published, not because it will enter the lake. No living organisms, no reagents, no field work are involved in Aim 1. What follows is what was actually built.
pFP-A encodes the mcjABCD operon for microcin J25 production — a 21-amino-acid lasso peptide active in the nanomolar range against Enterobacteriaceae, addressing fecal coliform contamination documented at 54× the regulatory limit in the Budi littoral zone (INDH 2024). pFP-B encodes a bicistronic SQR + PDO cassette for heterologous sulfide oxidation: SQR from Rhodobacter capsulatus, PDO from Cupriavidus pinatubonensis, with quantified sulfide-removing activity of 267 nmol·mg⁻¹·min⁻¹ at pH 8.0. pFP-C is the complete AND-gate plasmid: P_lux × P_phoB driving codon-optimized PhoA, with R-M shielding verified via REBASE + NEBcutter v3, and a temperature-sensitive origin of replication for seasonal consortium dissolution below 15 °C. All three strains are coupled through a circular auxotrophic ring (ΔhisD in A, ΔtrpB in B, ΔleuB in C).
The LuxR/AHL transfer function (K_d = 10 nM, n = 2) and the phosphate excess sensor (threshold 0.5 mg/L SRP — documented for Lake Budi by Quesille 2022) were modeled as Hill functions. The AND-gate truth table confirms PhoA output below 5% in all OFF states and above 95% in state (1,1) — validating digital-like conditional remediation. The AND-gate logic also appears in the Asimov Kernel simulation cross-validation.
A dimensionless Monod consumer-resource system for three strains and three metabolites was implemented in SciPy. The interior fixed point is asymptotically stable: all six eigenvalues of the symbolic Jacobian have Re(λ) < 0, dominant Re(λ_max) ≈ −0.215. The critical result is in the cascade perturbation: removing one member from the ring drives the remaining two to extinction within fewer than 70 hours, while removing one partner from a two-strain mutualism eliminates only one. That asymmetry — ring fails closed, pair fails open — is the topological biocontainment claim, and it is numerically established.
An escaped auxotrophic strain without metabolite supply decays exponentially, reaching functional extinction in under 24 hours. The evolutionary escape calculation correctly accounts for the generation-to-clock-time conversion factor g = μ_max / ln(2) — an easily missed dimensional error. The key qualifier: the ring does not improve escape time per se. The advantage is topological, not kinetic. Even if an escape mutant emerges, restoring independence from one partner still leaves it unable to receive metabolite from the third strain it can no longer access once the ring breaks.
What was discarded —
and why the project is stronger for it.
Füzi Poiesis did not begin as what it became. Several design directions were pursued, evaluated, and deliberately dropped. This section documents those decisions — not as failures, but as the methodology by which the final architecture was validated through elimination.
Three strains. Three functions. Unspecified organisms.
The project began as a systems-level design: three bacterial strains targeting three documented pathologies of Lake Budi, coupled through cross-auxotrophic dependencies. The logical architecture was sound from the beginning. What it lacked was biological specificity — the organisms were placeholders, and the DNA design had no sequences to anchor it.
This was the correct starting order of operations — build the mathematical framework first, then instantiate it — but it required a clear chassis commitment before any design work could proceed.
E. coli K-12 as the explicit demonstration chassis
The initial framing left organism identification as future work — contingent on metagenomic sampling of Lake Budi's sediment community. This was identified as a structural problem: a project cannot present DNA designs, ODE parameters, or Jacobian eigenvalues without a chassis whose sequences and kinetics are known. The metagenome of Lake Budi's benthic community does not exist in the published literature, and generating it would require years of field sampling and sequencing.
The resolution was clean: E. coli K-12 MG1655 was adopted as the demonstration chassis for Aim 1, explicitly named as such. Its published kinetic parameters (Shou et al. 2007; Mee et al. 2014) make the ODE model well-founded. The halotolerant progression — Halomonas elongata DSM 2581 → native Lake Budi isolates — is scoped to Aim 2, where it belongs. The distinction between "demonstration chassis" and "deployment organism" became a structural axis of the project's honesty.
Topology as biocontainment — no external kill switch
The early design included a MazE/MazF toxin-antitoxin system as an active genetic kill switch. Rottinghaus et al. (2022) demonstrated that single CRISPR-based kill switches fail in complex environments at predictable escape frequencies. Adding the MazE/MazF system on top of an auxotrophic ring would have introduced a complex, fragile maintenance requirement without meaningfully improving containment.
The design decision was to make the ring itself the biocontainment mechanism. The circular auxotrophic dependency — where the loss of any one member cascades to the extinction of all three — is a structural property that does not require active genetic maintenance. Complexity was removed; robustness increased. The ts-ori in pFP-C provides a complementary fail-safe through physical rather than genetic means: seasonal dissolution below 15 °C, matching Lake Budi's winter thermal regime.
Almost no limnology literature exists on Lake Budi
A systematic review of the available literature on Lake Budi revealed a significant institutional gap. Very few peer-reviewed studies address the lake's ecology in depth. The lake exists at the intersection of rural poverty, Mapuche territorial sovereignty, and low scientific visibility — and institutions had largely ignored it. The DGA's own commissioned review (LME-UChile 2010) confirmed this explicitly: hydrodynamics and trophic ecology were the least-represented research topics in the 77 documents surveyed.
This gap was initially a problem for the project's specificity. It became instead the project's most important contextual argument: the absence of data is itself a political fact, not a scientific one. The project was reframed to operate independently of the specific metagenome — proposing a modular consortium architecture that could be instantiated in any halotolerant chassis once native data exists, rather than waiting for data that may not arrive for years.
The empirical base was built piece by piece from the sources that do exist: Sandoval (2009) for halocline dynamics and benthic anoxia, Quesille (2022) for phosphorus concentrations and coliform documentation, LME-UChile (2010) for the socioeconomic baseline and trophic classification, Peña-Cortés (2006) for watershed anthropization, Stuardo et al. (1989) for the foundational physicochemical characterization. Each citation fills a specific evidential gap rather than serving as decorative bibliography.
The argument is coherent from beginning to end
Upon completion of the four Aim 1 deliverables, the project was reviewed against the criteria it had set for itself. The ODE model is well-formulated, with parameters sourced from the published literature. The hypothesis is falsifiable: if the interior fixed point loses stability when recalibrated with native strain parameters in Aim 2, the ring architecture requires redesign. The four computational panels answer exactly the four questions the project poses — monoculture decay, closed-pair stability, three-strain convergence, and cascade extinction contrast.
The AND-gate meets both quantitative criteria: below 5% output in all OFF states, above 90% in the ON state. The Jacobian gives asymptotic stability. Cascade collapse occurs in the time the theory predicts. What is presented for May 13 is a computational proof of concept with sufficient mathematical rigor — not a promise that the lake will be cleaned, but a demonstration that the topological architecture of the three-strain ring possesses properties that two-strain systems do not. That is exactly what Aim 1 claims to demonstrate, and it demonstrates it.
Three slides. One argument. One lake.
The project was presented in three slides: the topological biocontainment diagram showing the three-strain hypercycle against the eutrophication stratification of Lake Budi; the Aim 1 computational proof of concept — ODE modeling, Jacobian eigenvalue spectrum, AND-gate truth table, and the three annotated Benchling plasmid maps; and Aims 2 and 3 — from Halomonas elongata wet-lab validation through community-governed littoral deployment. Presented alongside co-author Benjamín Arias Almeida, HTGAA node coordinator.
"What is presented is a computational proof of concept with sufficient rigor. Not promising the lake will be cleaned — demonstrating that the topological architecture of the ring has mathematical properties two-strain systems don't possess. That is exactly what Aim 1 says it will demonstrate, and it does."
The boundaries of
what Aim 1 can claim.
These are not failures — they are the explicit falsifiability criteria that make Aim 2 meaningful. A project that documents its own limits is more trustworthy than one that doesn't acknowledge them.
The Monod kinetic parameters are established for E. coli K-12 under amino acid restriction, but have not been measured for halotolerant bacteria under Budi's salinity and temperature conditions. If the ODE model recalibrated with Halomonas elongata parameters loses its stable interior fixed point, the ring architecture requires redesign before Aim 2 can proceed. This is the primary falsifiability criterion for the transition between aims.
The Monod model assumes a well-mixed environment. In the lake, amino acid diffusion between strains is limited by physical distance. The model may overestimate coupling strength. A spatially explicit reaction-diffusion model is required to determine the critical inter-strain separation distance compatible with hypercycle stability — scoped to Aim 2, where the pumicite scaffold geometry must be evaluated against this constraint.
The evolutionary escape calculation uses a simplified single-locus model. Horizontal gene transfer from native lake organisms carrying intact leuB, trpB, or hisD homologs cannot be quantified without the Lake Budi native metagenome — which does not exist in the published literature. Chromosomal integration of all auxotrophic deletions, rather than plasmid-based delivery, is the primary mitigation strategy and a design requirement for Aim 2.
PhoB/PhoR in E. coli K-12 responds to phosphate deficiency, not excess. The Aim 1 X₂ transfer function is therefore an abstract validation of AND-gate Boolean logic — not an implementable molecular sensor in the demonstration chassis. The molecular design of an SRP-excess sensor is deferred to Aim 2 and depends on chassis selection.
The full metagenomics pipeline — native strain isolation from Lake Budi sediments, WGS methylation characterization, auxotrophic chassis domestication, and community-governed field sampling under Lafkenche FPIC — constitutes a postgraduate research agenda estimated at five to eight years. Aim 1 provides the mathematical foundation from which that work can begin with rigor rather than intuition. The Bokashi del Budi initiative at UFRO, currently in fabrication with an active budget of 400,000 CLP, is the parallel empirical track already in motion.
Golden Gate Assembly
Golden Gate Assembly uses Type IIS restriction enzymes — such as BsaI — that cut outside their recognition sequence, generating custom 4-bp overhangs that define the order and orientation of fragment joining. Unlike Gibson Assembly, which requires 20–40 bp overlaps and a 5′ exonuclease, Golden Gate enables scarless, ordered, one-pot assembly of multiple fragments in a single isothermal reaction.
This is directly relevant to Füzi Poiesis at the cassette level. The modular architecture of each plasmid — promoter → RBS → CDS → terminator, with annotated part boundaries in Benchling — is compatible with Golden Gate assembly logic for Aim 2, where multiple functional cassettes must be joined without scar sequences that could disrupt expression or introduce unintended open reading frames. The R-M shielding applied in Sub-Aim 1.1, which removes internal recognition sites via synonymous codon substitutions, would extend to Type IIS sites in a Golden Gate-compatible design. This recitation confirmed that the Aim 1 modular cassette structure is already aligned with the assembly strategy most likely to be used in Aim 2 wet-lab execution.