Engineering: Design → Build → Test → Learn
We engineered an aptamer-based system and companion software to iteratively improve our design across four DBTL cycles. Each cycle documents our Design, Build, Test, and Learn.
Cycle 1
Placeholder structure — add your Cycle 1 content.
Overview
TODO.
Design
TODO.
Build
TODO.
Test
TODO.
Learn
TODO.
Cycle 2
Placeholder structure — add your Cycle 2 content.
Overview
TODO.
Design
TODO.
Build
TODO.
Test
TODO.
Learn
TODO.
Cycle 3
Placeholder structure — add your Cycle 3 content.
Overview
TODO.
Design
TODO.
Build
TODO.
Test
TODO.
Learn
TODO.
Cycle 4
Overview
We wanted to create a plasmid backbone with aptamer arms that can bind to target molecules in aqueous solution. To do this, we used the pUC19 plasmid that has a LacZ promoter. We placed our aptamer-based vector under the control of the LacZ promoter in pUC19 where, under the presence of lactose (pUC19), the insert will create our lambda exonuclease and restriction enzyme, and, in the presence of glucose, will be inactivated for proper transport of our deliverable.
Design
We created an aptamer-anti-aptamer structure: the vector insert on the sense strand contains a complementary sequence to the anti-aptamer, which makes the anti-aptamer appear on the antisense strand. Therefore, once the plasmid is cut, the exposed 5’ end would be that of the anti-aptamer rather than the aptamer, which gets digested by the lambda exonuclease.
Key components:
- Spacer sequences: Selected with NOODL to avoid hairpins and entanglement.
- KpnI cut site: 5’-GGTACC-3’ centered to create aptamer arms after cleavage.
- Restriction endonuclease CDS: Under LacZ promoter to cleave the central site.
- Lambda exonuclease CDS: Under LacZ with HisTag; digests exposed 5’ ends (anti-aptamer).
- Rumble Zone: Pause sites flanking the insert to protect the backbone.
We used Golden Gate Assembly with BsaI (5’-GGTCTC-3’) modeled in Geneious. gBlock flanked by BsaI; fragments: vector (gBlock + Ultramer) and pUC19 backbone.
Build
PCR-added BsaI sites to ultramer, gBlock, and backbone; multistep Golden Gate at 1:2:2. Transformed DH5α via electroporation or Mix&Go; plated on LB+Amp (37 °C overnight).
Test
Colony PCR with plasmid- and gene-specific primers (NEB OneTaq 2X; ten 50 µL reactions; 30 s initial denaturation). Visualized with 1% agarose.
Learn
Design
Reused colonies and restreaked the index plate. Extended colony PCR initial denaturation to 5 min at 94 °C; used plasmid-specific primers only.
Build
Restreaked LB+Amp plates (37 °C overnight). New Golden Gate as in V1.0; transformed into fresh DH5α; incubated overnight.
Test
Visualized amplicons on 1% agarose.
Learn
Design
We needed to test whether LacI/LacIq clamp would prevent leaky λ-exo. DH5α chromosomal LacI activity was unclear, so we designed a blue-white screen.
Build
DH5α + native pUC19 on LB+Amp+X-Gal (±IPTG), 37 °C overnight.
Test
Observed mixed blue/white colonies on both X-Gal only and X-Gal + IPTG plates.
Learn
Software: NOODL
Overview
NOODL (Novel Optimization Of DNA Linkers) selects spacer sequences that avoid unwanted interactions and folding while preserving aptamer structure and accessibility.
- Generates spacer candidates using secondary-structure heuristics
- Filters by GC content, repeats, homopolymers, and restriction sites
- Exports sequences ready for cloning/ordering and Golden Gate assembly
Design
We set out to design single-stranded DNA spacer/linker sequences that maintain desired flexibility via (i) target A/T content, (ii) avoiding complementarity to aptamer sequences and plasmid flanks, (iii) minimizing specific motifs (such as palindromes, repeats, folding), and (iv) respecting fixed-end constraints. The key idea was to use a genetic algorithm (GA) to search the sequence under a multi-term scoring function where a lower score is better. Fitness is derived as fitness = max(score) - score for selection. An important requirement for our application, driven by our wet lab team’s procedures, was to strictly avoid interaction between spacer and its flanking regions. Our scoring function therefore penalizes any predicted hybridization. Since these interactions can inhibit toxin-aptamer binding, this is essential to prevent the invalidation of experimental outcomes.
Build:
Code modules include:
- Crossover: Implements multiple types of crossing over methods (Multipoint, Single, Uniform). Selects Multipoint crossover as default, if the user did not input a selection
- RCScore: computes composite penalties; utilities for reverse complement search
- BleedingFlanks: builds boundary k-mer sets to span junctions between fixed and variable nucleotides from flank content; scores complement hits
- BiasSelection: implements multiple types of bias (Stochastic Universal Sampling, Roulette, and Tournament). Selects Tournament Bias as the default if the user did not input a selection
Run Sequence:
- Create random sequences based on user input
- Compute kmer dictionary for randomly generated sequences and flanking regions; compute the reverse complements of everything in that dictionary
- Evaluate score
- Select parents using bias method; apply crossover/mutation; enforce penalties
- Score Re-evaluation to determine single best spacer sequence
Test
We took multiple sequences produced by NOODL and ran them through UNAFold’s DNA Folding Form as well as Geneious’ DNA folding tool. In those tools we also observed a folding difference. Based on the differences observed, we refined NOODL’s input parameters. The goal was to identify the parameters that consistently produced 1) a low internal score and 2) desirable folding
Our main changes in testing included:
- BiasSelection: experimented with different kinds of bias to see which types of bias gave us the best scoring sequences
- Crossover Functions: experimented with different crossing over points in the sequence to see which point gave us the best scoring sequences
- Mutation Rate: experimented with values between 0.05 and 0.2
- Population Size: experimented with values between 100-300
- Number of Generations: The size of the k-mers we are counting, which ranged from 2-10
- Size of the k-mers we are counting: ranged from 2-10
When we found that the UNAfold server was down, we completely pivoted to modeling on Geneious’ DNA folding tool. While exploring the software, we found that changing the temperature of the folding simulation at testing at 20°C, 37°C, and 55°C had a significant impact on the spacer’s predicted thermodynamics. At higher temperatures, the spacer exhibited a greater (ΔG), indicating lower stability and resulting in greater structural flexibility. On the other hand, at lower temperatures, the spacer is predicted to be thermodynamically rigid and stable.
Learn
What we observed:
- Tournament selection increased population diversity and reduced repeated parent picks within a selection pass, leading to steadier convergence than roulette well or stochastic universal sampling
- The reverse complementary kmer dictionary strategy of scoring works well, as when we’re counting kmers after plotting the structure of final sequences, we observe the correct amount of kmers
- Bias Selection: Tournament
- Crossover Type: Multipoint
- Mutation Rate: 0.2
- k-mer Length: 3-6
Ideal input parameters:
Towards the final days of our project, we discovered that the p4g03 aptamer was the cause of experimental inconsistency. The p4g03 was being tested without its overhang sequence while the p6 aptamer was correctly tested with its overhang. This difference led to the visible degradation of the p4g03 aptamer by the spacer sequence.
After learning this piece of information, our team ran NOODL with inputs that accurately modeled both p6 and p4g03 with its overhang. Our program generated a spacer that was shown to protect the structural integrity of both aptamers.
This outcome refined our initial hypothesis. We previously assumed the p4g03 aptamer was being degraded solely because it had a higher kd value than p6 (the lower the kd the more structurally sound the aptamer would be). However, we discovered that a critical experimental factor, the absence of p4g03’s overhang, was the primary driver of instability.
NOODL (Novel Optimization Of DNA Linkers) selects spacer sequences that avoid unwanted interactions and folding while preserving aptamer structure and accessibility.
- Generates spacer candidates using secondary-structure heuristics.
- Filters by GC content, repeats, homopolymers, and restriction sites.
- Exports sequences ready for cloning/ordering and Golden Gate assembly.
See more on our Software page.