Week 1 HW: LungLite — Principles, Practices, and Governance

🌬️ Project Idea: LungLite (AI + Breath Microfluidics + Cell-Free Synbio)

1) Biological engineering application/tool + why

LungLite is a low-cost, noninvasive breath monitoring system that uses a microfluidic disposable cartridge.

The cartridge contains freeze-dried cell-free synthetic biology reactions to detect breath biomarkers associated with airway inflammation and oxidative stress.

A smartphone camera reads the cartridge’s color/fluorescence pattern and an AI model interprets the result.

The tool is intended to help users monitor lung health over time—especially people with asthma, COPD risk, and high pollution exposure—and provide early warning signals of inflammation before severe symptoms appear.

LungLite leverages cell-free synthetic biology to detect breath biomarkers safely and efficiently. Instead of using live engineered cells, it employs freeze-dried transcription-translation (TX-TL) systems with non-replicating DNA circuits that respond to molecules associated with airway inflammation and oxidative stress. When a user exhales into the microfluidic cartridge, these engineered circuits trigger colorimetric or fluorescent signals proportional to biomarker levels. The sealed cartridge design, combined with built-in post-reaction neutralization, ensures safety, while AI algorithms analyze the visual output to provide an accurate, real-time readout of lung health. This integration of synthetic biology, microfluidics, and AI enables a low-cost, noninvasive tool for continuous monitoring, especially in high-risk environments or populations with limited access to traditional respiratory diagnostics.

Why this matters:
Current lung monitoring tools like spirometers often require strong forced exhalation and are not always accessible, comfortable, or usable for children, elderly people, or individuals in low-resource settings.

This problem is also deeply personal to me because I grew up around severe air pollution in Delhi, where “bad air days” are normal and respiratory symptoms are common. LungLite is motivated by the idea that people in high-exposure environments should be able to track early signs of inflammation easily and affordably—before symptoms become severe.

Initially worked on an AI-powered diagnostic tool for lung cancer. During this opportunity, I pivoted the design to focus on the Present Idea: a low-cost, noninvasive breath test that uses a microfluidic cartridge to track early signs of lung inflammation.

LungLite goal:
breathe → cartridge reacts → phone reads

References-

Cell free systems:

https://pmc.ncbi.nlm.nih.gov/articles/PMC11920963/
https://www.mdpi.com/1422-0067/25/16/9109
https://www.nature.com/articles/s41467-021-25233-y
https://www.mdpi.com/1420-3049/29/8/1878
Biochemical Preparation of Cell Extract for Cell-Free Protein Synthesis without Physical Disruption

DNA Circuits:

2) Governance/policy goals for an ethical future

Because LungLite sits at the intersection of bioengineering + consumer health + AI, it raises issues in biosecurity, lab safety, privacy, equity, and responsible health claims.

The governance goal is to ensure LungLite contributes to an ethical future by preventing harm while promoting constructive public health benefits.

Policy Goal A — Enhance Biosecurity

Sub-goal A1: Prevent incidents
Prevent misuse of cartridge biology (DNA templates, cell-free reagents) for harmful applications.
Sub-goal A2: Help respond
Ensure traceability and safe reporting if unsafe use or distribution occurs.

Policy Goal B — Foster Lab Safety

Sub-goal B1: Prevent incidents
Ensure safe handling, manufacturing, and disposal of cartridges and reagents.
Sub-goal B2: Help respond
Ensure protocols exist for spills, exposure, or improper disposal.

Policy Goal C — Protect the Environment

Sub-goal C1: Prevent incidents
Ensure cartridges and reagents do not introduce living organisms into waste streams.
Sub-goal C2: Help respond
Ensure recall, disposal, and remediation pathways if materials are found to persist or contaminate waste streams.

Policy Goal D — Other considerations

Minimize costs and burdens to stakeholders
Ensure feasibility for student prototyping and future scaling
Do not unnecessarily impede legitimate research
Promote constructive applications (public health monitoring, pollution health impacts)

3) Governance actions

Option 1: Technical Safety-by-Design

(Cell-free only + built-in kill chemistry + non-replicating DNA templates)

Idea

Many biosensors rely on living engineered organisms or wet reagents that could survive handling errors. LungLite instead commits to a cell-free-only architecture, using non-replicating DNA and post-reaction neutralization so the cartridge cannot become a biological propagation risk.

Design

Actors: student researchers, academic labs, cartridge designers, manufacturers.

Key elements:

Use commercially available or lab-prepared TX-TL cell-free extract
Use DNA templates without replication machinery
Add nuclease or denaturing reagents in a sealed “waste chamber” that activates after the reaction
Design the cartridge as a sealed unit so users cannot access wet reagents directly
Provide clear disposal instructions (trash-safe, not drain)
Include a QR code for standardized disposal instructions and recall notices

Assumptions

Cell-free systems are safe enough for consumer-adjacent use
DNA templates cannot be easily repurposed into harmful functions
Cartridge sealing prevents tampering and accidental exposure
Neutralization chemistry is robust across temperature/humidity variation

Risks of failure

Users could physically open the cartridge, mishandle reagents, or bypass neutralization
Poor sealing could cause leakage
DNA templates could be shared and repurposed outside intended use

Risks of “success”

Widespread adoption could normalize at-home “bio reaction kits” without safety literacy
Overconfidence in “bio-safe” claims could reduce careful oversight and institutional review

Option 2: Distribution + Supply Chain Controls

(DNA sequence screening + controlled reagent distribution + batch traceability)

Purpose

Even if the platform is designed safely, misuse risk increases when synbio components are distributed widely. This option adds governance at the distribution layer, aiming to prevent malicious acquisition or repurposing of DNA templates and reagents.

Design

Actors: DNA synthesis companies, cartridge manufacturers, distributors, university procurement offices, and potentially regulators.

Key elements:

DNA template sequences are screened using existing industry DNA synthesis screening norms
Cartridges sold with batch numbers, manufacturer ID, and basic traceability
Reagent supply chain restricted to verified vendors

Assumptions

Screening reliably catches harmful sequences
Vendors cooperate and screening is consistently implemented
Traceability meaningfully deters malicious use
Legitimate users will tolerate additional friction

Risks of failure

DIY synthesis or black-market sources bypass screening
Screening could generate false positives and slow benign development
Increased cost and friction could reduce adoption in low-resource communities

Risks of “success”

Centralization of power in a small number of vendors could limit open science
Smaller labs, students, and global south researchers could be excluded due to cost and access barriers
Overly broad screening could suppress legitimate respiratory health research

Option 3: Responsible Health Claims + Data Governance

(Limit medical claims + privacy-by-design + transparency)

Aim

Even if the biology is safe, LungLite could still cause harm through false reassurance, panic, biased AI outputs, or privacy breaches. This option focuses on preventing digital harms and misleading health interpretation.

Design

Actors: app developers, product companies, IRBs/ethics boards (if research), privacy regulators, public health agencies, and clinical collaborators.

Key elements:

Position LungLite initially as wellness monitoring, not a medical diagnostic
Focus on trend tracking rather than absolute disease classification
Provide clear disclaimers (“not a diagnosis; seek medical care if symptoms worsen”)
Use local-first processing: results computed on-device when possible
Require informed consent for any cloud upload or model improvement
Provide opt-out for data sharing
Publish model limitations and performance across demographics
Align product claims with existing regulatory distinctions between wellness tools and regulated diagnostic devices

Assumptions

Users understand “monitoring” vs “diagnosis”
Privacy measures meaningfully reduce harm
AI transparency improves trust and responsible use
The model will generalize across different phones, lighting, and populations

Risks of failure

Users may treat outputs as diagnoses and delay care
Data leaks could expose sensitive health data
Model bias could cause false reassurance or false alarms in specific groups
Smartphone hardware variability could distort readings

Risks of “success”

A widely adopted breath-health dataset could become commercially valuable and exploited
Insurers/employers/schools could pressure people to share breath scores (coercive screening)
“Wellness” framing could still function as a de facto diagnostic

4) Scoring matrix (1 = best, 3 = worst; n/a allowed)

Does the option:	Option 1	Option 2	Option 3
Enhance Biosecurity
• By preventing incidents	1	1	2
• By helping respond	2	1	2
Foster Lab Safety
• By preventing incident	1	2	2
• By helping respond	2	2	2
Protect the environment
• By preventing incidents	1	2	2
• By helping respond	2	2	2
Other considerations
• Minimizing costs and burdens to stakeholders	1	3	2
• Feasibility?	1	2	1
• Not impede research	1	3	1
• Promote constructive applications	1	2	1

Figure: LungLite governance scoring matrix

Scoring justification:

Option 1 reduces biological risk at the source and does not rely heavily on enforcement.
Option 2 is strongest on biosecurity response, but worst on cost, equity, and research openness.
Option 3 is strongest for AI/privacy harms but does not fully address upstream biosecurity.

There are few environmental concerns regarding this device like: packaging waste at scale, there might be low environmental risk regarding cell-free extracts, small risks associated with chemicals and dyes. Mitigation can be: minimal-material design, sealed leak-proof cartridge, and take-back/clinic disposal at scale.

5) Prioritized strategy

Recommended strategy

I believe we should prioritize Option 1 + Option 3 as the core approach now, and adopt a lightweight version of Option 2 only once scaling and commercialization begins.

Why Option 1 is essential

Option 1 addresses the biggest safety and biosecurity concern,i.e, distributing engineered biological systems into homes. By committing to cell-free synthetic biology only, LungLite becomes safer, easier to dispose of, and easier to govern ethically.

Why Option 3 is equally critical

Even if the biology is safe, LungLite can still cause harm through:

false reassurance
panic from false positives
privacy breaches
biased AI outputs

Option 3 reduces these risks through responsible messaging, careful AI design, and privacy-by-design.

Where Option 2 fits

Option 2 becomes more important once LungLite is manufactured at scale. Heavy supply chain restrictions too early could:

block student prototyping
increase costs
reduce equitable access
slow research innovation

So the staged approach is:

Option 1 + Option 3 now
Option 2 later (commercialization / mass distribution)

Tradeoffs considered

Safety vs accessibility
Innovation vs security
User empowerment vs medical risk
Privacy vs model improvement

Audience for recommendation

This governance strategy is best targeted at:

MIT/university lab leadership
future consumer product manufacturers
public health agencies
privacy regulators

6) What I Learned

Ethical concerns that arose

Dual-use risk
AI harm
Privacy
Equity
Regulatory gray zone
Coercion risk (monitoring becomes surveillance)

Governance actions proposed to address these

Use cell-free systems only and avoid living organisms
Seal cartridges and neutralize biological material post-test
Implement privacy-by-design + local-first processing
Avoid medical claims until clinically validated
Keep manufacturing scalable and affordable
Add anti-coercion safeguards (minimize retention, discourage third-party access)

Week 2 Lecture Prep

Homework Questions — Professor Jacobson

1) DNA polymerase error rate, genome comparison, and how biology handles the discrepancy

Nature’s machinery for copying DNA is DNA polymerase. High-fidelity replicative DNA polymerases (with proofreading) have an error rate of approximately:

~1 error per 1,000,000 to 10,000,000 base pairs

Comparison to the human genome

The human genome is approximately:

~3,200,000,000 base pairs

If replication relied only on polymerase accuracy:

At 1 error per 1,000,000 bp:
3,200,000,000 / 1,000,000 = 3,200 errors per genome replication
At 1 error per 10,000,000 bp:
3,200,000,000 / 10,000,000 = 320 errors per genome replication

So even “high-fidelity” polymerase alone would still introduce hundreds to thousands of mistakes each time the genome is copied.

DNA polymerase’s shape precisely fits correct base pairs and uses a conformational “proofreading” motion to minimize misincorporation. https://www.sciencedirect.com/science/article/pii/S0969212615002695

How biology deals with the discrepancy

Biology reduces the final mutation rate using multiple layers of error correction:

Polymerase proofreading removes many misincorporated bases during replication.
Mismatch repair (MMR) fixes errors missed by proofreading.
Base excision repair (BER) fixes chemically damaged bases.
Nucleotide excision repair (NER) removes bulky lesions.

Together, these systems reduce the effective mutation rate to roughly:

~1 error per 1,000,000,000 to 10,000,000,000 bp per cell division

That means across one human genome replication, the final result is typically on the order of:

~0.3 to 3 mutations per cell division

🧪 Homework Questions — Dr. LeProust

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

The most commonly used method is:

Solid-phase phosphoramidite DNA synthesis

This is the standard chemistry used by most commercial oligo suppliers. It works by building a DNA strand one nucleotide at a time on a solid support (like a bead, column, or array surface) using repeated cycles of:

deprotection
coupling
capping
oxidation

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

Direct synthesis becomes difficult past ~200 nucleotides because:

A) The yield drops exponentially with length

Each synthesis step has less than 100% efficiency, so errors compound as the oligo gets longer.Even in an optimistic scenario, most strands are truncated or incorrect.

B) Errors accumulate

Long oligos contain more:

deletions (from incomplete coupling)
substitutions (from incorrect incorporation)
depurination damage (especially A/G under acidic conditions)
truncated fragments

C) Purification becomes difficult and expensive

Separating a perfect 200-mer from 199-mer and 198-mer fragments is hard, so cost and complexity increase quickly.

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

Because the yield would collapse to essentially zero and the error rate would be unusable.

A) Yield becomes extremely low

B) The error rate becomes unacceptable

Even the rare full-length molecules would almost always contain:

substitutions
deletions
truncations
damaged bases

So you would not get a clean, correct 2000 bp product.

What is done instead in practice?

Instead of direct synthesis, genes are made by:

synthesizing shorter oligos (usually 60–200 nt)
assembling them into longer DNA using methods like:
- Gibson Assembly
- PCR-based assembly
- Golden Gate
- Ligase Cycling Assembly (LCA)
then sequence-verifying clones to find a correct one

📄 HW by Dr. George Church — Grant Application (Devised)

Project Title

LungLite: A Room-Temperature, Breath-to-Color Microfluidic Cartridge Powered by Cell-Free Synthetic Biology and Smartphone AI for At-Home Lung Inflammation Monitoring

1) Abstract

Chronic respiratory disease affects hundreds of millions globally, yet lung health monitoring remains clinic-centered, effort-dependent, and inaccessible for many populations. Existing tools such as spirometry require strong forced exhalation and proper technique, while lab tests for inflammation and oxidative stress are expensive and slow.

I propose LungLite, a low-cost breath monitoring system that combines breath condensation microfluidics, freeze-dried cell-free synthetic biology, and smartphone computer vision + AI. Users breathe into a disposable cartridge that captures breath condensate and routes it through multiple reaction zones. Each zone contains a freeze-dried cell-free reaction that produces a colorimetric/fluorescent signal in response to oxidative stress and inflammation-associated breath chemistry.

A smartphone reader standardizes illumination, quantifies reaction outputs, and uses machine learning to interpret a multi-zone “fingerprint” into a trend score. LungLite is designed for safe, scalable, room-temperature storage and distribution and aims to enable daily lung health monitoring outside specialized medical centers.

2) Specific Aims

Aim 1 — Engineer a breath-to-fluid microfluidic cartridge
Hypothesis: A passive, low-cost cartridge can consistently convert breath into a defined liquid sample volume and deliver it to reaction zones with minimal variability.
Outcome: consistent fluid delivery across users and breathing conditions.

Aim 2 — Develop a multi-zone freeze-dried cell-free synbio sensing panel
Hypothesis: freeze-dried cell-free reactions can be stabilized at room temperature and produce reproducible outputs when rehydrated.
Outcome: 6–12 zone panel with internal controls and reproducible readouts.

Aim 3 — Build a smartphone reader + AI pipeline
Hypothesis: smartphone imaging + AI normalization improves reliability and interpretability.
Outcome: trend score + confidence + invalid-test detection.

3) Significance

LungLite could enable:

noninvasive monitoring
high-frequency measurement
accessibility for children and low-resource settings
room-temperature distribution
population-level monitoring during wildfire smoke events

4) Innovation

Cell-free synbio in a consumer cartridge
Fingerprint sensing rather than single biomarker
AI as a reliability layer (normalization + invalid detection + confidence)

5) Technical Approach and Work Plan (12 months)

Months 1–2: breath capture + condensation
Months 2–4: routing + zone array
Months 3–7: freeze-dry stabilization
Months 5–8: phone reader + illumination
Months 7–10: AI training + invalid detection
Months 10–12: validation + usability

6) Expected Deliverables

disposable cartridge (6–12 zones)
freeze-dried reaction panel + controls
smartphone reader dock
AI pipeline
validation report
product pathway plan

7) Risk Analysis and Mitigation

biomarkers variable → fingerprint + controls + AI
stability issues → sealed packaging + desiccant
diagnostic misuse → wellness framing + disclaimers
privacy misuse → local-first + opt-in + deletion

8) Safety, Ethics, and Governance Plan

cell-free only
sealed cartridges
built-in neutralization
sequence screening at synthesis
traceability if scaling begins
bias testing + transparency
no disease claims until validated

9) Team and Resources

Cross-disciplinary team spanning:

microfluidics
cell-free synbio
optics + computer vision
ML
product design

10) Long-Term Vision and Commercialization

reusable reader + disposable cartridges
room-temperature shipping
low-cost manufacturing (paper microfluidics)
Year 1: wellness monitoring
Year 2+: clinical validation + regulated pathway

HW Review Papers — Week Summary Notes

1) DNA Sequencing at 40 (Shendure, J., Balasubramanian, S., Church, G. et al. https://doi.org/10.1038/nature24286)

Idea

DNA sequencing has gone through multiple revolutions and now functions as a universal molecular measurement tool — not just a way to read genomes.

Key points

In ~40 years, sequencing scaled from kilobases → first human genome → millions of genomes
Sequencing is no longer only for genomes; it is now used to measure:
- gene expression (RNA-seq)
- chromatin state (ATAC-seq, ChIP-seq)
- lineage tracing
- somatic mutations
- molecular interactions
Costs dropped dramatically due to next-generation sequencing (NGS)
Authors argue sequencing’s long-term impact may rival the microscope

Key message

We have become extremely good at reading DNA at massive scale, speed, and low cost.

2) DNA Synthesis Technologies to Close the Gene Writing Gap (2023), Hoose, A., Vellacott, R., Storch, M. et al. https://doi.org/10.1038/s41570-022-00456-9

Focus

We still cannot write DNA as efficiently as we can read it — and this is a major bottleneck for synthetic biology.

Key points

Synthetic DNA is essential for:
- synthetic biology
- gene therapy
- DNA data storage
- nanotechnology
Current chemical synthesis struggles beyond ~200 base pairs
Long DNA synthesis is expensive and error-prone
New approaches aiming to scale DNA writing include:
- enzymatic (template-independent) synthesis
- microarray-based synthesis + assembly
- rolling circle amplification
- molecular assembly + cloning pipelines
As DNA writing becomes easier, regulation and oversight become more important

3) Recombineering and MAGE (2021), Wannier T, et al. Nat Rev Methods Primers, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9083505/

Core idea

Recombineering and MAGE enable precise, scarless, multiplex genome editing without requiring toxic double-strand breaks (DSBs).

Why traditional editing is limiting

Older editing methods (ZFNs, TALENs, CRISPR with DSBs):

rely on double-strand breaks
DSBs can be toxic (especially in bacteria)
repair often produces unwanted indels
low precision for large-scale combinatorial editing

Recombineering solution

Uses phage proteins (Redβ, Exo, Gam)
Introduces ssDNA or dsDNA with homology
DNA integrates at the replication fork
No DSB required
Editing is highly precise and “scarless”

MAGE (Multiplex Automated Genome Engineering)

Introduces many ssDNA oligos at once
Creates combinatorial diversity across many genomic sites
Enables genome-scale reverse genetics

4) CRISPR Technology: A Decade of Genome Editing is Only the Beginning, Wang, Doudna, et al., https://www.science.org/doi/10.1126/science.add8643

Focus area

CRISPR made genome editing programmable, accessible, and fast — dramatically lowering the barrier to entry.

Main points

Cas9 + guide RNA enables targeting by base pairing
Enabled:
- knockouts
- pooled genetic screens
- animal models
- crop editing
- emerging human therapies

Newer CRISPR-derived tools

Base editing: A→G or C→T without DSBs
Prime editing: templated edits with higher precision

Remaining challenges

off-target effects
delivery into cells/tissues
limited multiplexing at large scale
HDR inefficiency in many systems

Summary

Biotechnology has made DNA reading extremely scalable (sequencing), but DNA writing (synthesis) and DNA rewriting (editing) are still constrained by cost, accuracy, delivery, and scalability.

Sequencing is now a general-purpose measurement tool, while synthesis and editing are rapidly improving — raising both exciting capabilities and new governance needs.

I used artificial intelligence tools, including ChatGPT-5.0, for language refinement, structural organization, and clarity of expression in this documentation. All scientific concepts, design decisions, sequence selections, experimental reasoning, and technical interpretations reflect my own understanding and work. The AI tool was used solely to improve readability, coherence, and presentation quality.