Multimodal Assistive Communication System for Nonverbal Users Proposed Bioengineering Application I plan to focus on developing a multimodal assistive communication system for nonverbal users that integrates speech, gestures, and biological signals such as muscle activity or eye movement. The goal of this system is to help individuals with communication disorders express themselves more accurately and autonomously by leveraging whichever input methods are most reliable for them. The reason I want to go through with this topic is because I personally believe that it closely connects to my interests in Human-Computer Interaction and is a project where the concepts of Bioengineering would highly benefit it.
Part 1: Benchling and In-Silico Gel Art I made a Benchling accountand imported the Lambda DNA genbank sequence from https://www.ncbi.nlm.nih.gov/nuccore/J02459.1?report=genbank&log$=seqview After importing everything I used Benchling’s Digest tool to stimulate the enzyme digestion of the Lambda DNA using the required enzymes. I first made all of the enzymes that were asked in the list.
Homework Questions from Professor Jacobson Nature’s machinery for copying DNA is called polymerase. What is the error rate of polymerase? How does this compare to the length of the human genome. How does biology deal with that discrepancy?
The error rate of DNA polymerase with proofreading has an error rate of approximately 1 error per 10⁶ base pairs copied (10⁻⁶ per base). The human genome is ~3.2 billion base pairs (3.2 × 10⁹ bp). If polymerase made errors at 1 in 10⁶ bases, then you would expect roughly ~3,200 errors per genome replication. Biology uses multiple layers of error correction:
Python Script for Opentrons Artwork from opentrons import types import math
metadata = { ‘author’: ‘Shubhadip Saha’, ‘protocolName’: ‘H+S’, ‘description’: ‘A heart with the letters H+S inside of it.’, ‘source’: ‘HTGAA 2026 Opentrons Lab’, ‘apiLevel’: ‘2.20’ }
TIP_RACK_DECK_SLOT = 9 COLORS_DECK_SLOT = 6 AGAR_DECK_SLOT = 5 PIPETTE_STARTING_TIP_WELL = ‘A1’
well_colors = { ‘A1’ : ‘Red’, ‘B1’ : ‘Green’, ‘C1’ : ‘Orange’ }
Subsections of Homework
Week 1 HW: Principles and Practices
Multimodal Assistive Communication System for Nonverbal Users
Proposed Bioengineering Application
I plan to focus on developing a multimodal assistive communication system for nonverbal users that integrates speech, gestures, and biological signals such as muscle activity or eye movement. The goal of this system is to help individuals with communication disorders express themselves more accurately and autonomously by leveraging whichever input methods are most reliable for them. The reason I want to go through with this topic is because I personally believe that it closely connects to my interests in Human-Computer Interaction and is a project where the concepts of Bioengineering would highly benefit it.
Ethical and Governance Goals
A few key goals from what could be achieved from this project are:
Protecting user autonomy and agency, the system should reflect the user’s true intent, and should not override control over what is trying to be communicated.
Non-malfeasance and safety, minimizing harm that could arise from misinterpretation of biological signals or misuse in high-stakes environments such as hospitals.
A third goal is promoting equity and accessibility, ensuring that the system works across diverse bodies, abilities, and socioeconomic contexts.
These major goals could also be broken down into smaller sub-goals that could go over consent, bias reduction safeguards and more.
Governance Actions
1. Clinical Use Certification (Rule / Requirement)
The primary actors for this would be federal health or regulatory agencies, hospital procurement committees and medical boards
Purpose: Many assistive communication tools are currently released as consumer devices or research prototypes and are sometimes adopted informally in clinical settings without consistent validation. So, a good formal clinical-use certifications for such systems would be useful before they can be used in a high-stakes environment like a hospital ER or ICU. The certification would have to focus on safety, reliability and transparency.
Design: A standard would be developed specifically for communication-assistive systems, and it should include performance benchmarks and some form of precedence and proof of functionality from use in simulated clinical testing. We would have to get federal regulators to define and enforce this standard. Hospitals would have to receive certifications for procurement.
Assumptions: This assumes that regulators can design a proportionate certificate pathway that doesn’t slow innovation, and also that vendors and researchers would be able to afford validation costs, and that hospitals would be able to enforce the procurement standards.
Risks: If certification is too burdensome then nobody would get it. Slow regulatory timelines might push clinicians to use uncertified tools. Certification may also bring about a false sense of safety, possibly leading institutions to neglect training.
The primary actors for this would be hospitals, research institutions, disability advocacy organizations.
Purpose: Consent practices for communication technologies can vary widely, and some rely heavily on proxy consent that don’t require observable user assent or confirmation. Standardized informed consent processes should be implemented that ensure that the system reflects the user’s intent.
Design: Institutions would adopt consent workflows adapted to user abilities, such as eye tracking or simple binary signals. We would need confirmation tests that show runtime confirmation, cancellation or correction mechanisms. Hospitals and other organizations should enforce these standards, and vendors should build compliant interfaces.
Assumptions: This is only possible assuming that the user of this device is capable of providing some sort of reliable assent signal, and also that institutions treat consent as an ongoing process rather than a yes or no question. (covered in the next governance action)
Risks: Consent standards may fail if users with highly variable signals cannot reliably confirm an output, which can potentially restrict access to the technology.
3. Safeguards for Ambiguity and Uncertainty (technical strategy)
The primary actors for this would be developers, research labs, companies.
Purpose: Uncertainty is often hidden from users and clinicians. Many assistive systems present signal interpretations as definitive outputs, even when they are sometimes noisy. Systems should be able to explicitly manage uncertainty by requesting confirmation when confidence of a decision is low.
Design: Systems would implement calibrated confidence metrics, and adjustable thresholds based on context, and clear indicators of uncertainty in a decision. Developers would have to build, labs to validate calibrations, and other organizations can support in standardization.
Assumptions: Assumes that confidence estimates can be calibrated, and that users and clinicians can interpret uncertain information correctly.
Risks: Poor calibration can easily mislead users and cause bad decisions. Excessive uncertain information can erode trust and slow communication.
4. Equity and Bias Auditing (standards)
Purpose: Assistive communication systems are often trained on narrow datasets, leading to uneven performance across users. We need to establish regular, standardized audits that evaluate system performance across diverse bodies, abilities, and contexts.
Design: Audits would implement test datasets, and other metrics to quantify a system’s ability. Independent audits could also conduct evaluations.
Assumptions: Assumes diverse and representative datasets can be assembled responsibly, and that vendors approve of third-party evaluations.
Risks: Audits may still be able to miss underrepresented groups or incentivize optimization for benchmarks rather than real results.
Table
Does the option:
Option 1
Option 2
Option 3
Option 4
Enhance Biosecurity
n/a
n/a
n/a
n/a
• By preventing incidents
n/a
n/a
n/a
n/a
• By helping respond
n/a
n/a
n/a
n/a
Foster Lab Safety
2
n/a
1
2
• By preventing incident
2
n/a
1
2
• By helping respond
2
n/a
1
2
Protect the environment
n/a
n/a
n/a
n/a
• By preventing incidents
n/a
n/a
n/a
n/a
• By helping respond
n/a
n/a
n/a
n/a
Other considerations
• Minimizing costs and burdens to stakeholders
3
2
2
3
• Feasibility?
2
1
2
2
• Not impede research
2
1
2
2
• Promote constructive applications
2
1
1
1
Prioritization
I would prioritize the Clinical Use Certification and the Safeguards for Ambiguity and Uncertainty as the most important governance actions for guiding the development and deployment of the multimodal assistive communication system. I believe that achieving these two goals would most directly address the risks of harm that arise when implementing these systems into a medical context.
Certifications may introduce additional costs and slower deployment, but its a price that that can be paid to guarantee a safer product that can provide better communication for more efficient uses of medical resources, and can also prevent serious medical, legal or ethical consequences.
Safeguards for ambiguity and uncertainty are equally important, as they address the confusions that can be caused by the interpretation of biological signals. This governance action reduces the probablility of miscommunications and prevents systems from overstepping user intent. It helps to maintain trust and safety.
The other governance goals are important complementary measures, but without strong safety measures and uncertainty-aware system design, consent mechanisms and audits alone may be insufficient to prevent harm.
After importing everything I used Benchling’s Digest tool to stimulate the enzyme digestion of the Lambda DNA using the required enzymes. I first made all of the enzymes that were asked in the list.
EcoRI
HindIII
BamHI
KpnI
EcoRV
SacI
SalI
EcoRV made a lot of small fragments, which made very dense lower bands, BamHI produced less mid-size fragments.
At first I was experimenting with the fragments to make a “:D” smiley face. But later, while working through it I realized that I had created more of a wine glass / goblet looking shape and, hence, I pivoted to making just that.
For the final composition, I arranged the following enzyme digests across lanes:
Lane 1: BamHI
Lane 2: EcoRI
Lane 3: HindIII
Lane 4: EcoRI
Lane 5: BamHI
The symmetric outer lanes create the bowl width, the EcoRI lanes taper inward, and the HindIII digest forms the vertical stem due to its smaller fragment sizes. When cropped and viewed holistically, the gel bands form the silhouette of a wine glass.
Part 3: DNA Design Challenge
I picked Ovalbumin as my protein. The reason I picked this protein is becuase it is the main protein that is used in egg whites, and it’s historically important in immunology. (and i like eggs)
The optimized ovalbumin gene was cloned into a pET plasmid for expression in E. coli. After transformation into bacterial cells, the T7 promoter drives transcription of the DNA into mRNA. Ribosomes translate the mRNA into ovalbumin protein. The protein can then be purified from bacterial lysate. This process follows the central dogma: DNA is transcribed into RNA, which is translated into protein.
Part 4: Prepare a DNA Synthesis Order
4.2
For plasmid assembly, I selected the pTwist Amp High Copy vector, which is designed for high-copy replication in E. coli and contains an ampicillin resistance marker for selection. This backbone supports efficient bacterial transformation and protein expression. The codon-optimized ovalbumin expression cassette was inserted into this circular backbone, resulting in a 6504 bp plasmid construct.
Part 5: DNA Read
(i) What DNA would you like to sequence and why?
I’d like to sequence synthetics DNA used for digital data storage. I remember we briefly went over this topic in the last lecture, but I found that it was very interesting that data could be stored in such a way. DNA-based storage encodes binary information into nucleotide sequences. This sequencing can allow the retrieval of the stored data by reading the encoded bases and reconstructing the digital information. I’d like to do this specifically because feel that it bridges biology and computation, and it demonstrates how DNA can be a high-density storage medium beyond its biological role.
(ii) What sequencing technology would you use?
Illumina is a second generation sequencing technology that offers veru high accuracy and massive parallel sequencing, which means it is well suited for decoding short synthetic DNA fragments used for digital storage systems.
The output consists of millions of short DNA reads in FASTQ format, containing both sequence information and quality scores. These reads can be computationally assembled to reconstruct the original stored data.
5.2 DNA Write
(i) What DNA would you want to synthesize and why?
Long-term archival data storage DNA strands. Digital files to binary code to nucleotide sequencers through mapping schemes. This offers very high data density, and long-term stability compared to traditional hard drives.
(ii) What technology or technologies would you use to perform this DNA synthesis and why?
DNA Phosoramidite Synthesis. Nucleotides are added sequentially to a growing DNA strand attached to a solid support.
5.3 DNA Edit
(i) What DNA would you want to edit and why?
TERT gene in human somatic cells plays a role in cellular aging. Telomeres shorten during cell division. Understanding and modulating TERT expression can provide insights into aging, longevity and age-related diseases.
(ii) What technology would you use?
CRISPR-based base editing to modify the TERT gene (telomerase reverse transcriptase) in human somatic cells. Telomeres shorten with age, contributing to cellular senescence, and carefully regulating telomerase activity may help improve regenerative capacity. Base editing allows precise single-nucleotide changes without creating double-strand breaks, reducing the risk of large mutations. A Cas9 nickase fused to a deaminase enzyme, guided by RNA, converts specific bases within a targeted region. However, limitations include off-target edits and the risk of increased cancer if telomerase activity becomes uncontrolled.
Week 2 Lecture Prep Questions
Homework Questions from Professor Jacobson
Nature’s machinery for copying DNA is called polymerase. What is the error rate of polymerase? How does this compare to the length of the human genome. How does biology deal with that discrepancy?
The error rate of DNA polymerase with proofreading has an error rate of approximately 1 error per 10⁶ base pairs copied (10⁻⁶ per base). The human genome is ~3.2 billion base pairs (3.2 × 10⁹ bp). If polymerase made errors at 1 in 10⁶ bases, then you would expect roughly ~3,200 errors per genome replication. Biology uses multiple layers of error correction:
Proofreading by DNA polymerase during replication
Post-replication mismatch repair systems (e.g., MutS/MutL pathways)
Redundancy in the genome (non-coding regions, diploidy)
Cellular quality control, including apoptosis for heavily damaged cells
Together, these reduce the effective mutation rate to ~10⁻⁸ per base per generation.
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?
DNA has 64 codons encoding 20 amino acids, so most amino acids are encoded by multiple codons. An average human protein is ~1036 bp (~345 amino acids). Because many amino acids have 2–6 synonymous codons, the number of possible DNA sequences that encode the same protein is extremely large. (on the order of 10^100+ possible sequences)
In practice, most of these possible codes do not work effectively. This is due to factors such as codon bias (preferred codons differ between organisms), mRNA secondary structure that can inhibit transcription or translation, unintended regulatory signals embedded in DNA, repetitive or GC-rich sequences that are difficult to synthesize, and effects on translation speed that influence protein folding. As a result, although many sequences are theoretically valid, only a small subset function well in real biological systems.
Homework Questions from Dr. LeProust
What’s the most commonly used method for oligo synthesis currently?
The most commonly used method for oligonucleotide synthesis is chemical phosphoramidite synthesis on a solid support. This method adds nucleotides one at a time through repeated cycles of coupling, capping, oxidation, and deprotection. It is highly automated, scalable, and reliable for short oligos, which is why it remains the industry standard.
Why is it difficult to make oligos longer than ~200 nt via direct synthesis?
Each phosphoramidite synthesis cycle has a small but nonzero failure rate. As the oligo gets longer, these errors accumulate exponentially, leading to truncated products and sequence errors. Longer sequences also suffer from increased depurination, incomplete coupling, secondary structure formation, and purification challenges. As a result, yield and fidelity drop sharply beyond ~200 nucleotides.
Why can’t you make a 2000 bp gene via direct oligo synthesis?
Direct chemical synthesis of a 2000 bp sequence would accumulate too many errors and truncations to be practical, resulting in extremely low yield of full-length, correct product. The chemistry lacks error correction, and the probability of obtaining a perfect 2000 bp strand is effectively near zero. Instead, long genes are made by assembling many shorter oligos (typically 60–200 nt) using enzymatic methods such as PCR-based assembly, Gibson assembly, or ligase-based assembly, followed by sequencing and error correction.
Homework Questions from 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”?
Using standard biological knowledge (what animals cannot synthesize de novo and must obtain from diet), the 10 essential amino acids in animals are:
Histidine (H)
Isoleucine (I)
Leucine (L)
Lysine (K)
Methionine (M)
Phenylalanine (F)
Threonine (T)
Tryptophan (W)
Valine (V)
Arginine (R)
Week 3 HW: Lab Automations
Python Script for Opentrons Artwork
from opentrons import types
import math
metadata = {
‘author’: ‘Shubhadip Saha’,
‘protocolName’: ‘H+S’,
‘description’: ‘A heart with the letters H+S inside of it.’,
‘source’: ‘HTGAA 2026 Opentrons Lab’,
‘apiLevel’: ‘2.20’
}
def location_of_color(color_string):
for well,color in well_colors.items():
if color.lower() == color_string.lower():
return color_plate[well]
raise ValueError(f"No well found with color {color_string}")
max_vol = 20
for i in range(0, len(red_points), max_vol):
batch = red_points[i:i + max_vol]
pipette_20ul.aspirate(len(batch), location_of_color(‘Red’))
for x, y in batch:
loc = center_location.move(types.Point(x=x, y=y))
dispense_and_detach(pipette_20ul, 1, loc)
pipette_20ul.drop_tip()
for i in range(0, len(h_coords), max_vol):
batch = h_coords[i:i + max_vol]
pipette_20ul.aspirate(len(batch), location_of_color(‘Green’))
for x, y in batch:
loc = center_location.move(types.Point(x=x, y=y))
dispense_and_detach(pipette_20ul, 1, loc)
pipette_20ul.drop_tip()
for i in range(0, len(s_coords), max_vol):
batch = s_coords[i:i + max_vol]
pipette_20ul.aspirate(len(batch), location_of_color(‘Orange’))
for x, y in batch:
loc = center_location.move(types.Point(x=x, y=y))
dispense_and_detach(pipette_20ul, 1, loc)
pipette_20ul.drop_tip()
Gemini Prompts
I want to create a heart with the letters H+S inside of it
I’ve added the code to generate the heart shape with ‘H+S’ inside. I used simple geometric calculations for the heart and coordinates for the letters, ensuring each color (Red for the heart, Green and Orange for the letters) is handled correctly with tip management.
Please explain this error:
The error ‘Aspirating 32uL … is more than the pipette can hold (20uL)’ happens because your code tries to pull up 32 microliters of red liquid for the 32 points of the heart, but the pipette only holds 20 microliters. I have updated the code to handle the heart in two smaller batches so it fits the pipette’s capacity.
can you make the border of the heart more defined and close together, also the H+S is not centered inside of the heart, you can make the heart bigger if needed, but make sure that there is no space between the dots of the border of the heart
I’ve refined the heart to make the border dots touch, creating a solid-looking outline. I also scaled up the heart and adjusted the ‘H+S’ coordinates to ensure they are perfectly centered within the shape.
the heart looks good, now the letters inside of the heart have to be fixed. the dots inside of the H are very spaced out. the plus sign doesnt look like a plus sign but instead looks like a diamond, and the dots are also very spaced out, and the S doesnt look proper
I’ve increased the number of dots for the letters ‘H’, ‘+’, and ‘S’ to make them look solid and well-defined, just like the heart border. I also reshaped the ‘+’ to be a clear cross and refined the ‘S’ curves.
I have updated the coordinates for the letters to be much denser so they appear as solid lines. I also refined the ‘+’ sign to be a clear cross and adjusted the ‘S’ shape to ensure it looks proper and fits nicely inside the heart.
ok now the H looks fine, and the plus looks fine, its just that the S now has a disconnect between the top half and the bottom half. also can you make the plus sign red instead of green
I’ve fixed the disconnect in the letter ‘S’ by adjusting the start and end points of the curves. I also moved the plus sign to the Red color section so it matches the heart border.
Post-Lab Questions
1. DeRoo, J. B., Jones, A. A., Slaughter, C. K., Ahr, T. W., Stroup, S. M., Thompson, G. B., & Snow, C. D. (2025). Automation of protein crystallization scaleup via Opentrons-2 liquid handling. SLAS technology, 32, 100268.
The paper talks about how how a low-cost Opentrons OT-2 liquid handling robot can automate the setup of protein crystallization experiments. Protein crystallization is essential for determining molecular structures but traditionally requires careful, repetitive manual pipetting. The authors developed Python scripts to automate the preparation of sitting-drop crystallization trials across multi-well plates. They found that the automated workflow reduced labor, improved reproducibility, and produced crystallization results comparable to manual methods.
Part 2:
For my final project, I intend to develop a multimodal assistive communication and physiological monitoring system that integrates biological sensing with digital interfaces. Specifically, I aim to explore the use of a biological biosensor (e.g., a cell-free or microbial system responsive to sweat metabolites such as lactate or pH) as an auxiliary communication channel. The long-term goal is to investigate how biological signals can support symptom monitoring during medical treatment or provide physiological feedback for applications such as athletic performance tracking.
Automation would play a critical role in ensuring the reliability and reproducibility of the biological sensing component. Using the Opentrons OT-2, I would automate the preparation and calibration of the biosensor assays. This could include automated serial dilutions to generate standard curves, precise dispensing of cell-free reaction mixtures, and batch preparation of replicate assay wells to evaluate sensitivity and variability. Automation would allow systematic testing of different analyte concentrations and environmental conditions, ensuring consistent reagent volumes and minimizing human error.
For example, an automated workflow might involve dispensing varying concentrations of lactate standards into reaction wells, followed by consistent addition of reporter reagents to quantify fluorescence or colorimetric output. Conceptually, this could be structured as:
for concentration in standard_curve:
pipette.aspirate(volume, analyte_stock)
pipette.dispense(volume, reaction_well)
Beyond assay calibration, I would also design a 3D-printed cartridge holder compatible with the Opentrons deck to standardize placement of biosensor strips during testing. This would improve spatial consistency and enable repeatable experimental conditions.
Overall, automation would allow me to focus on the human-centered aspects of the project. Specifically, how users interpret and respond to biological feedback — while ensuring that the underlying biological measurements are consistent and scalable. By integrating bioengineering workflows with automated liquid handling, this project would bridge synthetic biology, assistive technology, and human-computer interaction.
