Wet Lab

Qualitative assays for silicate solubilization

Design	Silicates in the Martian regolith are majorly in the form of anorthosite, a calcium-rich plagioclase feldspar. It is essentially a calcium aluminosilicate. Through existing literature we figured out that silicate solubilizing bacteria show differential activity on different silicate minerals ^[1]. Hence, it was imperative to show that our bacteria of choice Pseudomonas fluorescens shows solubilizing activity specifically with calcium aluminosilicate. We modified the protocol used in ^[1], where they used Burnt and Rovira medium substituted with agar and 0.25% calcium aluminosilicate. Due to unavailability of the soil extract required for the Burnt and Rovira medium, we decided to use LB medium. We also decided to try a protocol using dextrose agar substituted with silicates ^[2].
Build	0.25 grams of calcium aluminosilicate was added as a slurry to LB agar and dextrose agar prior to pouring into plates. We found a heating-cooling-heating method that helped in effectively dispersing the insoluble silicate. A 1000x diluted overnight liquid culture was used for inoculation using a spread-plating method.
Test	As suggested in the protocol, the plates were incubated for 3-5 days. After a few days we noticed some color change around the colonies in the LB plates, the area around the colonies appeared less cloudy but the appearance of conclusive clearance zones couldn’t be established with certainty. No colonies grew on dextrose agar.
Learn	We learnt that our P. fluorescens strain cannot grow on dextrose agar. Going further, we decided to choose a method that could show more conclusive evidence of silicate solubilization.

Design	In bacteria, both insoluble phosphate and silicate solubilization have a common mechanism - organic acid production. We decided to use a modified version of NBRISSM - a defined differential liquid media used for screening silicon solubilizers ^[3] . This media has been designed to ensure little to no interference from phosphate solubilization which ensures no false positive results. The indicator bromocresol purple (BCP) is added to obtain qualitative results. This indicator changes color from purple to yellow at pH 5.2. The pH change is due to the formation of silicic acid which makes the media change color from purple to yellow. The original composition uses magnesium trisilicate as the silicate source. However, given our context, we found it best to replace it with calcium aluminosilicate.
Build	We prepared the media as per the defined composition and replaced magnesium trisilicate with an equal amount of calcium aluminosilicate. The pH was adjusted to 7 before autoclaving. Bacterial inoculations were made at approximately 1–2 × 10⁹ CFU ml⁻¹. Uninoculated media served as a bacterial control whereas inoculated media without Si source acted as silica control. The culture tubes were kept at 30℃ at 250 rpm.
Test	We observed the tubes every 24 hours for 7 days and noted the color change. The color started changing after 4 days and became bright yellow by the 7th day. The color in both the control tubes remained purple, indicating a lack of solubilization in the control tubes. See our results here.
Learn	The color change was purely due to bacteria’s ability to solubilize silicon. As no color change was observed in tubes containing bacteria but no silicon, the possibility of bacterial growth itself lowering the pH and thus causing a color change can be ruled out. We can conclude that P. fluorescens can solubilize calcium aluminosilicate. Going further, we decided to prove its solubilizing activity on the calcium aluminosilicate present in our Martian soil simulant.

Growth of bacteria in Martian soil simulant

Design	Before testing out P. fluorescens ’s ability to solubilize the calcium aluminosilicate present in the Martian soil simulant (shown below), we needed to first establish a way to grow the bacteria in an otherwise non-life supporting environment. Following inoculation in soil, we needed to measure growth in the soil as well.
Build	After discussions with Dr. K Chandraraj, an expert in soil microbiology, we decided to use a cell suspension in fresh LB media to directly pour onto the soil and observe the bacterial growth.
Test	Everyday for 10 days, 1 gm of soil was mixed with sterile water and centrifuged. The supernatant was used for counting CFU. We observed an oscillatory growth pattern over the course of 10 days. See our results here.
Learn	This result is consistent with the findings of previous work that has shown the wave-like growth pattern of metabolically active cells to be a natural phenomenon associated with the cycles of growth and death in soil microbial communities ^[4] . Thus, we have established a method to propagate P. fluorescens in the Martian soil simulant in a way that mimics the natural growth patterns of bacteria in soil.

Quantitative assay for silicate solubilization in soil

Design	After having established a way to grow the bacteria in the Martian soil simulant, it was time to test its ability to generate solubilized silicates from the simulant. We followed a colorimetric method for quantifying silicate ions in solution. This meant we needed to extract the solubilized silicates from soil before the analysis. The solubilized silicates in soil can be extracted using a CaCl₂ solution ^[5]. We chose this over other extractants because it was reported that CaCl₂ and distilled water extracted readily soluble Si whereas other extractants either dissolved some exchangeable Si or specifically adsorbed Si. After extraction, the obtained silicate solution is diluted 41-fold. This is done to inhibit the formation of silica polymers as monomeric silica can quickly form polymers in high concentrations ^[6]. The diluted silicate ions in solution can be colorimetrically quantified using the silico-molybdate method. In acidic conditions, soluble silica reacts with molybdate ions to form a yellow complex (heteropoly acid), the intensity of which is directly proportional to the concentration of silicate ions. This complex when reduced turns to a bright blue color, the absorbance of which can be detected by a spectrophotometer at 630 nm ^[7]. Phosphate ions create an interference in this test because they also form a yellow complex with molybdate. To eliminate phosphate interference, the solution is treated with tartaric acid.
Build	We extracted silicates from the soil on which bacterial growth had been characterized. We also extracted silicates from uninoculated soil (control). The extracted solutions, the blank as well as silica standards were then subjected to addition of concentrated HCl to make the environment acidic. This was followed by addition of molybdate and tartaric acid. Finally, a reducing solution containing sodium sulfite, sodium bisulfite and ANSA (1-amino-2-naphthol-4-sulfonic acid) was added to the solution.
Test	The absorbance for all the silica standards and the test sample was measured at 630 nm. We received very erratic values for standards which made it difficult to plot a linear curve.
Learn	After going back to literature, we learnt that simply creating an acidic environment is not enough , the pH strictly needs to be maintained below 2.5. Additionally, we realized that we couldn’t use distilled water; we had to use deionized water for the reaction ^[6].

Design	We decided to increase our reaction volumes to minimize errors from pipetting small volumes and measure the pH before proceeding with the reaction.
Build	We used a pH paper to make sure the pH was below 2.5 and used deionized water. Additionally, we made sure no glassware was used. All the other steps were followed as done previously.
Test	We again received erratic absorbance values that couldn’t be used for plotting a linear graph. Qualitatively, the absorbance of our test sample was more than control.
Learn	On consulting Dr. Deepa Khushalani, an expert in the chemistry of silicon compounds, we were told that silicate estimation in solution can be very tricky due to the high tendency of polymer formation of monomeric silicates. Moreover, the time delay between the formation of silicates by bacteria and our quantification experiment can be a limiting factor for correct estimation as silicates quickly form polymers. We were prompted to think of other correlation establishing experiments. Therefore, going ahead we decided that co-relating an increase in diatom growth rate with bacterial presence in soil would be the best way for us to show our proof of concept.

Preparation of BioBricks

Design	We decided to engineer a limonene synthesis pathway into our chassis Pseudomonas fluorescens. This bacteria naturally produces GPP (geranyl pyrophosphate) which can be acted upon by the enzyme 4(S)-limonene synthase to produce 4(S)-limonene. The aim of our engineering, thus, was to introduce a constitutive production of the enzyme 4(S)-limonene synthase. We chose the constitutive promoter BBa_J23100 and the RBS BBa_J428038 as they have been extensively used and well documented for reliable expression. Our coding sequence was 4(S)-limonene synthase and the terminator was chosen to be BBa_J428092. Our destination vector was pJUMP24-1A(sfGFP), a high copy number E. coli-Pseudomonas shuttle vector with an sfGFP dropout marker.
Build	All the BioBricks were taken from the iGEM Distribution Kit 2024’s well plates as described in the handbook. We transformed the BioBricks into competent E. coli DH5α cells, and grew them on appropriate antibiotic plates. We had negative control plates for each antibiotic on which untransformed bacteria were inoculated.
Test	There were colonies for all the BioBricks. The colonies containing our destination vector were green indicating the sfGFP production. We observed no colonies on control plates.
Learn	Our transformation was successful and the colonies could be used for miniprep to isolate the plasmids.

Golden Gate Assembly

Design	All of our parts were MoClo compatible and were flanked by BsaI sites. BsaI is an exonuclease and creates overhangs a few base pairs away from the restriction site. The overhangs generated after restriction digestion would be compatible with the adjacent part’s overhang, thus ensuring all the parts are ligated in the proper order, i.e., vector-promoter-RBS-CDS-terminator-vector. The image below shows the overhangs used. The destination vector carries an sfGFP cassette as a dropout marker. This cassette will be removed as a result of BsaI digestion and will be replaced by limonene synthase cassette.
Build	Golden Gate Assembly is a single tube reaction that involves addition of all the parts required, the vector, the exonuclease and ligase together. We followed the protocol as mentioned in our experiments page. The assembly product was transformed into competent E. coli DH5α cells. The cells were grown on 35 µg/ml kanamycin plates.
Test	The destination vector is the only plasmid in the reaction with kanamycin resistance, which ensures that only those cells which were transformed with assembled vectors or self ligated vectors grew. The vector had an sfGFP cassette prior to the assembly which means that all the colonies with self-ligated vectors would retain their sfGFP production and can be screened by briefly exposing the plate under UV light. We received no green fluorescent colonies on our low CFU plate and several fluorescent green as well as non-fluorescent colonies on the high CFU plate. See our results here.
Learn	The presence of non-fluorescent colonies on the kanamycin plates confirmed the success of our Golden Gate Assembly. We proceeded to carry out further confirmatory tests.

Confirmation of plasmid assembly

Design

The assembled plasmid with the limonene production cassette has a size of 5.5 kb. Whereas, the plasmid with the sfGFP construct we started from is 4.4 kb long. We decided to linearize both the plasmids and run on an agarose gel to confirm the increase in size. The XbaI restriction site, as shown below, was used for this purpose. Additionally, we subjected the assembled plasmid to digestion with both XbaI and BstBI as these sites flanked our entire construct. The resulting fragments should be 2 kb (the construct) and 3.5 kb (the remaining plasmid backbone).

Plasmid	Restriction enzymes	Expected length of fragments
Assembled plasmid	XbaI	5.5 kb
Assembled plasmid	XbaI + BstBI	2 kb and 3.5 kb
pJUMP24-1A(sfGFP)	XbaI	4.4 kb

Build

We set up a restriction digestion reaction following the protocol mentioned in our experiments page.

Test

After the digestion, all the products were subjected to gel electrophoresis. We received a band at 4.4 kb for linearized pre-assembly vector (pJUMP24-1A(sfGFP) and a 5.5 kb band for linearized post-assembly vector. The double digestion product yielded 2 bands, one at 2.0 kb and the other at 3.5 kb.

See our results here.

Learn

The results from gel electrophoresis were as expected, indicating successful assembly. We decided to go ahead and transform the assembled plasmid into P. fluorescens.

Transformation of P. fluorescens

Design	Pseudomonas species have previously been reported to be transformed genetically by conjugation, transfection ,chemical transformation, or electroporation. P. fluorescens has been reported to show successful transformation using the standard electroporation protocol used for bacteria ^[8]. Since Pseudomonas species are reported to show increased rates of transformation in the presence of MgCl₂ in the electroporation buffer ^[9], we decided to used the method of subsequent washing of the cells in magnesium electroporation buffer (MEB) to prepare electrocompetent P. fluorescens cells and carry out transformation according to the electroporation protocol mentioned in our experiments page.
Build	We made electrocompetent cells using the MEB buffer and stored them as glycerol stocks in -80°C. Before transformation, the cells were thawed on ice. We set up two transformation reactions, one with 150 ng of DNA and the other with 1000 ng of DNA. The cells for negative control were not subjected to electroporation. All the cells were inoculated on kanamycin plates.
Test	After 36 hours of incubation, we saw 6 colonies on one of the test plates. However, there were 2 colonies on the negative control plate as well. To rule out faulty preparation of antibiotic plates, we made a new kanamycin plate. We picked the colonies and streaked on the new kanamycin plate. All the colonies from the negative control plate grew on the new plate. This prompted us to attribute the colony formation to contamination.
Learn	After inspecting the colony morphology, it was apparent that we had contamination. We figured out that electroporation cuvettes could have been a source of contamination.

Design	Due to the contamination problem from our first iteration, we decided to use new cuvettes for our second iteration.
Build	We repeated all the steps as done before with extra precaution.
Test	We saw one colony on the transformant plate with 150 ng DNA. No colonies were observed for 1000 ng DNA. No colonies were observed on the negative control plate. See our results here.
Learn	Better transformation of bacteria with 150 ng DNA as opposed to transformation with 1000 ng DNA can be attributed to steric hindrance caused by higher amounts of DNA. Presence of no colonies on negative control plates confirms the success of our transformation.

Culturing diatoms

Design	The diatom strain P. tricornutum SAG 1090-1a is listed as a brackish strain by the Culture Collection of Algae at the University of Göttingen. We decided to grow the culture in the modified Mann and Myers medium with NaCl 10 gm/L, MgSO₄.8H₂O 2.4 gm/L, CaCl₂.2H₂O 0.6 gm/L, supplemented with a trace element solution 20 ml/L ^[10]. Phosphorus and nitrogen were supplied as potassium phosphate (0.2 gm/L) and ammonium bicarbonate (0.3 gm/L).
Build	The medium was prepared and pH was adjusted to 7.1-7.8. It was inoculated in axenic conditions from the agar slant.
Test	The diatom culture showed increased turbidity but absence of brown pigment, indicating a deficit in photosynthetic activity. As an attempt to rescue the culture, we supplemented it with vitamin solutions: thiamine hydrochloride and cyanocobalamin, but there was no improvement.
Learn	We realized that vitamins were not the growth restricting factor and tried looking for other growth media.

Design	We found an alternative growth media for P. tricornutum, the BG11 medium ^[11]. Since it is a marine diatom, we decided to inoculate two flasks of BG11 media, prepared in distilled water and 100% natural seawater (with salinity 30%) respectively. We also decided to decrease the incubation temperature since microalgae have an optimum growth temperature at 20-25°C, with P. tricornutum showing maximum growth rate at 21°C.
Build	We inoculated the two media in axenic conditions and incubated at 21°C and 150 rpm.
Test	Growth was quantified by measuring the optical density of the culture at 750 nm. We observed ideal growth in the medium prepared in natural seawater.
Learn	The diatoms showed optimum growth when cultured in nutrient enriched natural seawater. Increasing the temperature further aided in the enhancement of growth rates, with 27°C showing slower growth but a similar saturation OD. Based on these results, we finalized the culture media as BG11 prepared in seawater and growth conditions as 27°C, 160 rpm.

Design	A few months into culture maintenance, the diatoms started showing a decline in growth. We tried to investigate various growth affecting factors. To determine the effect of varying salinity we decided to substitute the seawater with 25% and 35% saltwater. Since microalgae have an optimum growth temperature at 20-25°C, with P. tricornutum showing maximum growth rate at 21°C,we chose to adjust the temperature to 21°C.
Build	We inoculated two BG11 media with varying salinities and incubated them at 21°C, 150 rpm.
Test	No growth was observed in either of the media. We tried readjusting the pH of the media, inoculating in f/2 medium and reinoculating both BG11 medium and f/2 medium from the primary culture, with no results in any of attempts.
Learn	None of the changes showed any improvement in diatom survival, indicating that salinity, pH or composition of the growth media were not the restricting factors for diatom growth.

Design	We found that P. tricornutum can be a facultative heterotroph and utilize carbon source for better growth ^[12]. Based on this, we decided to supplement the diatoms with a carbon source in order to rescue the growth.
Build	We inoculated three sets of media: f/2 medium supplemented with glycerol at a concentration of 0.1M, f/2 medium supplemented by glucose at a concentration of 10 gm/l, and the BG11 medium, and incubated at 21°C to accelerate the growth.
Test	The diatoms showed growth in the f/2 + glycerol medium and slight growth in BG11 medium. These cultures when shifted to 27°C continued showing stable growth. See our results here.
Learn	Supplementing the diatoms with a carbon source might help in bypassing certain metabolic pathways, thus rescuing its growth. In co-culture conditions, the lysate of dead bacteria can act as the carbon source for the diatoms and enhance growth as well as increase survival under stressful conditions.

Minimal media culturing

Design	To facilitate long term space exploration by means of reducing payload, we are required to minimize the resources utilized for our co-culture setup. In order to minimize the resource requirement for diatom culturing, we optimized the most minimal media required for P. tricornutum growth. We decided to supplement the diatoms with only the essential nutrients: nitrogen and phosphorus, and eliminated all trace nutrients and vitamins ^[13].
Build	The media was prepared by supplementing natural seawater with nitrate and phosphate in the form of sodium nitrate (75 mg/L) and sodium dihydrogen phosphate (5.65 mg/L), according to the concentrations of the salts used in the f/2 medium ^[13]. Growth conditions were maintained at 27°C and 160 rpm.
Test	Growth quantification results for 15 days showed a standard growth curve with exponential phase occurring between days 5 and 8. See our results here.
Learn	This experiment successfully demonstrated that P. tricornutum can thrive in a nutrient-deficient media containing solely nitrogen and phosphorus sources.

Effect of silicon on diatom growth

Design	Silicate solubilizing bacteria act by converting the silica present in minerals to silicic acid, which can be utilized by the diatoms in cell wall formation ^[14]. Sodium silicates are one of the few water soluble silicates, and form silicic acid at neutral pH conditions. We test the hypothesis that the diatoms show increased growth in presence of silicic acid in the growth medium by comparing their growth in presence and absence of sodium metasilicate.
Build	We harvested diatom cells by centrifugation of equal volumes of the stock culture at 8000 rpm for 10 min, and resuspended them in the two media. We ensured the same growth conditions for all the cultures, which was 27°C, 160 rpm and 1800 lux light intensity.
Test	We observed similar growth rates in presence and absence of silicate in the exponential phase. However, as the diatoms approached stationary phase, the cultures containing silicate showed a consistently increasing OD as compared to cultures without silicate. See our results here.
Learn	Presence of silicate in the media leads to higher overall growth of P. tricornutum, confirming that the diatoms can take up solubilized silica from the media for cell wall formation, resulting in growth enhancement.

Preparation of genes for acetaminophen synthesis

Design	We sought out to demonstrate the dual biomanufacturing capabilities of our co-culture system by engineering the diatoms to synthesize acetaminophen. We received the acetaminophen synthesis genes, 4abh and nhoA, from Twist Bioscience in 2023. We received the P. tricornutum-specific plasmids pPTbsr and pPtPuc3 from Addgene as stab cultures. After propagating the culture on appropriate antibiotics we made glycerol stocks and isolated plasmids from the culture. We decided to insert our genes of interest in the multiple cloning site flanked by the fcpA promoter (a constitutive promoter) and terminator.
Build	We designed Gibson assembly primers for 4abh and nhoA genes for acetaminophen synthesis. We set up a PCR reaction for amplifying the 4abh and nhoA genes.
Test	We did not receive any bands. We tried again with a higher concentration of the template DNA and received very faint bands.
Learn	It was reasoned that the template DNA had degraded due to non-optimal storage. To test this we performed an electrophoresis on just the template DNA and received a smear, which confirmed our suspicion. Since we were facing issues with our diatom’s growth at the same time, we decided not to proceed with transforming them.

Optimizing salinity for co-culture

Design	P. tricornutum, a marine diatom, is shown to require a saline environment for optimum growth. The diatom culturing experiments indicated an optimum salinity of 30-35%, which corresponds to media prepared in 100% seawater. Pseudomonas species, including P. fluorescens, are typically grown in LB media having lower salt concentrations. To determine the maximum salinity of the minimal media that can be used for co-culturing, we designed a setup with the martian soil simulant and cultures of P. fluorescens inoculated in minimal media of varying salinities.
Build	We inoculated P. fluorescens in the Martian soil simulant according to the protocol optimized in our previous experiments. 3 gm samples of the inoculated soil simulant were added to four Falcon tubes each containing around 15 ml of minimal media prepared with salinities of 30%, 15%, 7.5% and 5% and incubated at 30°C. The cultures were grown for 4 days and the growth was determined by counting the CFU of each sample.
Test	Samples from the four cultures were plated on agar and incubated overnight at 30°C to determine CFU of each culture. The CFU readings were taken twice in a period of four days both of which showed dense colony formation in all four samples. See our results here.
Learn	The results show that the salinity of media used in the co-culture setup does not hinder the growth of P. fluorescens as they formed dense colonies when plated. Therefore, we can carry out co-culturing of P. fluorescens with P. tricornutum in media having the salinity levels optimum for diatom growth, which is 30-35%, corresponding to 100% seawater.

Testing the co-culture

Design	Diatoms are commonly found in symbiotic relationships with bacteria and other organisms in their natural habitat. These relationships are mutually beneficial to the organisms by virtue of establishment of nutrient cycles. We designed a co-culture between P. fluorescens and P. tricornutum which exhibits a silicate solubilization and uptake cycle along with carbon and micronutrients cycling. One of the important factors for a successful co-culture is optimal starting inoculation ratio that accounts for the differences in growth rates of the organisms. We chose 1:50 (bacteria: diatoms) as our inoculation ratio. We decided to quantify the growth by measuring OD of the cultures at two different wavelengths: 600 nm and 750 nm. This was chosen since bacteria exhibit an absorbance maxima at 600 nm and diatoms at 750 nm.
Build	We measured the concentration of cells in the subcultures of both the organisms by cell counting using a haemocytometer and accordingly harvested cells in a ratio of 1:50 for bacteria:diatoms. This ratio was chosen after accounting for the difference in doubling rates of the two species. The co-culture was set up along with three controls to compare growth of the organisms in co-culture with the organisms alone in the same environment.
Test	We quantified growth of the two organisms in the co-culture setup by measuring optical density at 600 nm and 750 nm, for bacteria and diatoms respectively. We observe that the bacteria and diatoms in the co-culture show synchronous growth, with a wave-like pattern corresponding to growth and death cycles of the bacteria in soil. On comparing the growth of P. fluorescens and P. tricornutum in co-culture with the controls, we observe that both the organisms show a higher average growth in the co-culture environment as opposed to when grown separately. See our results here.
Learn	The synchronicity of the diatom and bacterial growth implies a co-dependence of the two species for resources and indicates the establishment of a nutrient cycle. The bacteria can serve as a carbon source for enhancing the growth of the diatoms and in return, the diatoms, upon death, act as a source of silica for the bacteria. This observed interdependence and an increase in the average growth rates of the two organisms shows the establishment of a mutualistic relation between P. fluorescens and P. tricornutum in our co-culture setup.

Model

Nitrate in Diatom Media

Design	The initial metabolic model for Phaeodactylum tricornutum was designed based on a medium containing very low traces of nitrate. We assumed that this would not significantly affect the metabolic fluxes and proceeded with this assumption in our analyses, focusing on pathways related to biomass and acetaminophen production.
Build	Using this model, we generated various plots and performed a series of analyses, making several inferences about key metabolic pathways, particularly those involved in biomass synthesis and acetaminophen production.
Test	However, we found that the growth rates were not consistent with what we were observing in the wet lab.
Learn	After conducting a thorough literature review, we realized that nitrate content plays a critical role in influencing metabolic fluxes, including those central to both biomass and acetaminophen production.

Design	This discrepancy between the low-nitrate model and the actual conditions in wet lab experiments indicated the need for a more accurate representation of the medium.
Build	In response, we modified the model to simulate a medium with higher nitrate content, closely resembling the conditions used in wet lab experiments.
Test	After rerunning the analyses with this updated medium, we generated new plots.
Learn	This produced more accurate results and improved our understanding of the nitrate’s impact on the metabolic pathways.

Pseudomonas Modeling

Design	The initial metabolic model for Phaeodactylum tricornutum was designed based on a medium containing very low traces of nitrate. We assumed that this would not significantly affect the metabolic fluxes and proceeded with this assumption in our analyses, focusing on pathways related to biomass and acetaminophen production.
Build	Using this model, we generated various plots and performed a series of analyses, making several inferences about key metabolic pathways, particularly those involved in biomass synthesis and acetaminophen production.
Test	However, we found that the growth rates were not consistent with what we were observing in the wet lab.
Learn	After conducting a thorough literature review, we realized that nitrate content plays a critical role in influencing metabolic fluxes, including those central to both biomass and acetaminophen production.

Design	To model the silicate solubilization by Pseudomonas fluorescens, we decided to use organic acid production as an indicator of solubilization. This was done based on our literature review into the mechanisms of solubilization.
Build	We attempted to incorporate a reaction for silicate solubilization involving gluconic acid into an existing , which lacks a GSMM. The initial plan was to modify the model by adding this reaction and analyzing its effects.
Test	On testing the model, we found that the model we used was incomplete, as the developer had not included the metabolite names, only their IDs. This became evident when we attempted to add the silicate solubilization reaction and could not identify gluconic acid.
Learn	To overcome this limitation, we searched for a more complete model that included all metabolite names.

Escher

Design	We aimed to visualize the modified diatom metabolic pathways by uploading the SBML file of the diatom’s GSMM to the Escher website. This tool was chosen for its ability to automatically display metabolic pathways based on the provided model, allowing us to analyze the modifications in a streamlined way.
Build	After uploading the SBML file, the website generated pathways that initially appeared accurate. However, upon closer inspection, we discovered that the pathways displayed were from E. coli rather than the diatom model.
Test	We attempted various troubleshooting methods to resolve the discrepancy but were unable to determine why the website defaulted to E. coli pathways, despite providing the correct diatom SBML file.
Learn	As a result, we turned to a manual approach to create the required visualizations.

Design	We found that the Escher Python library would allow us to create our own plots.
Build	We constructed the pathways using the Escher Python library.
Test	Although this process was labor-intensive, it ultimately yielded accurate pathway visualizations.
Learn	This allowed us to make key inferences about the diatom’s metabolic modifications.

SimBiology Network Development

Design	A network model for silicate solubilization was designed using organic acid production, silicon uptake, and solubilization occurred, assuming all processes happened within the cell.
Build	The initial model was built in SimBiology to simulate a single compartment where all reactions occurred.
Test	We ran tests using this model, but the silicate levels were significantly overestimated in the simulation.
Learn	Upon further literature review, we discovered that organic acids are secreted outside the cell, where the actual solubilization takes place.

Design	Our initial assumption of internal solubilization likely caused the misrepresentation of silicate values.
Build	To correct this, we switched to a two-compartment model that better reflects real-life conditions, distinguishing between intracellular processes and extracellular solubilization.
Test	We tested our new two-compartment model in SimBiology.
Learn	This resulted in more accurate silicate flux values.

SimBiology Rate Constants

Design	The first model incorporated filler rates based on preliminary literature review, leading to the use of rate values that were neither well-calculated estimates nor accurately sourced from previous studies.
Build	Recognizing this limitation, we focused on gathering more reliable data and incorporated rate constants for well-established reactions, such as gluconolactone oxidation.
Test	While testing the new model, we made informed estimates of missing parameters by analyzing the model’s output graphs and data.
Learn	Using the updated model, we identified certain reactions as enzyme-driven processes, applying Michaelis-Menten kinetics to enhance the model's accuracy. These improvements led to a more robust and reliable model, with accurately defined rate constants that better reflect the biochemical processes involved.

Goal of Dynamic Modeling

Design	Our initial project goal was to provide the wet lab with precise numerical values for the time required for silicon solubilization and the saturation point of solubilization, measured in mmol of silicic acid.
Build	We designed a model to simulate these processes based on available data and assumed we could make accurate, quantitative predictions.
Test	As we began constructing the simulation, we quickly encountered a data crunch - there was insufficient usable information available about certain pathways and processes related to silicate solubilization. This lack of data made it impossible to build a one-to-one simulation that accurately reflected real-world conditions.
Learn	In response, we shifted our focus from quantitative outputs to qualitative insights. Specifically, we designed experiments to investigate how solubilization changes with increasing pH and to determine how solubilization plateaus over time and at specific pH levels. This pivot allowed us to gather valuable qualitative insights into the solubilization process. We were able to enhance our understanding of the relationship between pH and silica solubilization, even without precise numerical data, by observing trends and plateaus in the solubilization process. This learning phase provided critical insights that can inform future experiments and model improvements.

Community Modeling

Design	We designed a simulation of the bacteria and diatoms in co-culture using the MICOM framework.
Build	We constructed a community model using the same external media as the diatom’s GSMM.
Test	The initial growth media used in the community model did not yield any fluxes for biomass, resulting in a growth rate of zero.
Learn	After conducting a literature review, we discovered that both the bacteria and diatom require specific molecules like magnesium and urea in the community model’s media to support their growth.

Design	We identified the molecules to be added to the community media.
Build	We incorporated these essential molecules into the growth media, anticipating that this change would enhance the model's performance.
Test	Following the adjustments, we observed non zero growth and fluxes for both the diatom and the bacteria.
Learn	This indicated that the inclusion of the necessary growth factors was successful in facilitating their metabolic activities.

Software

BioCyc

Design	We envisioned a tool that would take a ChEBI ID as input and return organisms capable of metabolizing the resource using data from BioCyc.
Build	We utilized Python's requests module and a third-party GitHub library to query BioCyc for metabolic pathways and organism data.
Test	The tool inconsistently returned results due to issues with the external library and BioCyc's access restrictions.
Learn	Relying on third-party libraries is inherently risky, and we needed a custom solution to avoid access limitations and improve reliability.

Design	We planned to replace the third-party library with a custom querying system and create a session-based login for BioCyc’s paid access.
Build	We developed our own Python-based solution to directly query BioCyc and established a session-based login to handle access restrictions.
Test	The custom solution consistently retrieved data and successfully returned organisms associated with the inputted ChEBI IDs.
Learn	Building a custom querying solution improved reliability, but further filtering was needed to focus on relevant chassis organisms. This also highlighted the need for proper error handling and made it clear that we had to provide a simpler to use interface than a command-line environment to widen usability for users of varying technical expertise.

Design	We aimed to refine the tool by filtering out non-microbial organisms, focusing on bacteria, archaea, and microbes suitable for synthetic biology.
Build	We implemented a filtering algorithm that excluded plants and animals by parsing taxonomic data for each organism.
Test	The refined tool returned only microbial chassis, improving the relevance of results for synthetic biology applications.
Learn	Improved filtering enhanced organism selection, and we recognized the opportunity to expand the tool to support product-based biomanufacturing in the future.

UniProt

Design	Started with understanding basic API querying in Python using requests. The goal was to learn how to send a GET request to UniProt's API with a specific query, retrieve JSON data, and handle basic filtering for reviewed bacterial organisms.
Build	Wrote a simple Python code to query the UniProt API using the base URL and a basic EC number or ChEBI compound index. Used requests.get() to retrieve data and json() to parse it.
Test	Run the code to check if the API request returns the desired data. With proper error handling for unsuccessful requests (e.g., wrong URLs, no internet connection).
Learn	Gained familiarity with the requests library, how to make API calls, and extract basic information from the response.

Design	Extended functionality to filter duplicate organisms by their scientific name, count occurrences, and sort by frequency. Aim to avoid multiple results for the same organism and focus on the most frequently appearing organisms.
Build	Implemented a Python function to filter duplicates based on scientific names, count occurrences, and store organisms in a dictionary.
Test	Confirmed that the function successfully removes duplicates, counts the number of occurrences, and sorts them by frequency.
Learn	Refined understanding of working with dictionaries and list comprehensions in Python. Developed logic to aggregate and sort data.

Design	Added more advanced functionality, such as exporting filtered results to a JSON file and only printing organisms with high annotation scores (≥ 3). The goal is to integrate data output for further downstream analysis.
Build	Used Python’s json library to write results to a file and filter based on annotation score.
Test	Verify that the filtered results are correctly exported as JSON and that the printed results match the annotation score condition.
Learn	Improve file handling, learn more about data serialization (JSON), and understand how to perform deeper filtering on the returned data.

KEGG and BRENDA

Design	Tasked with implementing compound querying via KEGG’s REST API, a new experience for us, we had to learn how to structure queries, communicate with web APIs in Python, and retrieve relevant data for the compounds we were interested in.
Build	Using Python, we implemented the API calls and successfully set up basic compound queries, which returned data on compound-enzyme relationships. We also explored how to use SMILES notation for querying compounds to bridge communication between multiple databases.
Test	The queries worked well for a limited set of compounds. However, the pathway to translate compounds to organisms was unclear. This led to delays in getting from compound to organism.
Learn	We realized we had to take a deeper dive into KEGG’s architecture to understand how its databases interact. Additionally, SMILES notation proved to be a valuable tool for cross-database compatibility, and we would leverage this knowledge later in the project.

Design	Initially, we were exploring the compound → enzyme → reaction pathway → organism approach. However, navigating KEGG’s BRITE hierarchy through its API became an obstacle due to the complexity of its reaction pathway structure.
Build	We shifted our focus to a new relationship chain: compound → enzyme → gene → organism. This was more streamlined and bypassed the problematic BRITE hierarchy. We implemented this new approach using KEGG’s API and tested it on a few compounds.
Test	While the new approach worked, it was too slow. Each compound query took approximately 4 minutes, which was unacceptable for the expected 10-20 compounds per run. Query bottlenecks became apparent during the gene-to-organism step, where KEGG’s limits on API calls significantly slowed progress.
Learn	We learned that while the compound → gene → organism pathway was optimal in terms of data retrieval, we needed to optimize our queries to improve performance. We also realized that BRITE’s complex hierarchy should be avoided, especially for large-scale queries.

Design	Faced with long query times, we aimed to significantly speed up the process. Our plan was to optimize the API interactions by reducing redundant queries and implementing parallel processing.
Build	We first refined our queries by batching them and eliminating duplicate calls. Then,we implemented multiprocessing to allow parallel querying. We also realized that KEGG’s gene codes were prefixed with organism codes, allowing us to bypass unnecessary gene-organism queries by creating a local lookup table of organism codes.
Test	The optimizations reduced query times from 4 minutes per compound to around a minute. Local searching allowed me to quickly match genes to organisms without repeated API calls, and multiprocessing shaved off additional time.
Learn	We learned the value of preprocessing and local storage for optimization. The ability to locally reference organism codes sped up the pipeline and significantly reduced query times. While KEGG’s query limits remained an obstacle, the combination of multiprocessing and local searching allowed us to mitigate it.

Design	To complement KEGG, we decided to query BRENDA for additional EC numbers. However, BRENDA’s SOAP API, which was unfamiliar to us, required an entirely different architecture compared to KEGG’s REST API. We set out to learn and implement this new API architecture.
Build	We implemented the BRENDA API using the Python zeep library. However, we encountered significant issues - many documented API calls were not functioning as expected, with filtering parameters not being processed correctly server-side. Despite copying the examples directly from the documentation, several calls returned incomplete or erroneous data.
Test	After multiple failed attempts, we abandoned the problematic API calls and developed a workaround using BRENDA’s CSV download functionality. We set up Python code to automate the downloading of search results as CSV files, adjusting the search fields directly in the download URL.
Learn	We learned that while SOAP API architecture could potentially provide more structured responses, its implementation on BRENDA’s side was incomplete or flawed. The CSV download method proved much more reliable, allowing us to seamlessly extract the necessary EC numbers without further API issues.

Scoring Function and API Integration

Design	With a list of candidate bacteria generated, We were tasked with developing a confidence score system. BacDive would be queried to check if the suggested bacteria are typically cultured in conditions similar to the user’s input.
Build	We implemented the BacDive API to retrieve temperature ranges, biosafety ratings, and (where available) media information for the bacteria. The process was straightforward since BacDive used REST architecture, but inconsistent documentation formats across bacteria entries caused frequent errors during data retrieval.
Test	While BacDive’s temperature and biosafety data were reliably accessible, media composition and pH values were not consistently provided. Additionally, media references often led to external papers, which we could not reasonably scrape within the project’s scope.
Learn	We learnt the importance of error handling in bioinformatics. The inconsistent formats in BacDive’s documentation led to broken loops in the code, which we resolved through extensive error handling. Though BacDive had limitations, it served as a crucial component for building the confidence score.

Design	While BacDive provided useful environmental data, many media references linked to MediaDive. We wanted to integrate MediaDive’s API into the pipeline to enhance the confidence score’s accuracy by obtaining more reliable pH and media composition data.
Build	We implemented MediaDive’s API, which had an extensive and well-documented REST interface. The code was designed to query MediaDive whenever BacDive referenced culture media from that database, pulling in more detailed media composition and pH information.
Test	The integration with MediaDive significantly improved the confidence score. The availability of additional media data allowed for a more thorough comparison between the user’s input conditions and the documented conditions for bacterial growth.
Learn	Integrating MediaDive made the confidence score system more robust, offering more precise matches between input parameters and bacterial conditions. This collaboration between BacDive and MediaDive provided a more accurate and reliable final output for the user.

Django Frontend Design and Integration

Design	The goal was to create a user-friendly search form for querying the ChEBI database to create a streamline and self-contained user experience. This also had to be usable across browser types and versions. Implement responsive and visually appealing forms using Bootstrap as the CSS framework.
Build	Implemented the search form in the search.html template, including input fields for the search term, pH range, temperature range, and conditions using Django classes. Utilized sessionStorage to store user inputs dynamically within the script elements of the html, on client side.
Test	Conducted usability testing with different browsers and devices to ensure the search form behaved as expected. Validated that the input fields accept correct data formats and handle edge cases (e.g., negative pH).
Learn	Gained a basic familiarity of the Django framework. Obtained compatibility information of the frontend with various web browsers.

Design	Set out to create a boiler-plate for CHEBI API requests structure set-up. Wanted to determine the necessary data points to extract required information from the ChEBI API.
Build	Implemented the search_chebi function in views.py in order to query the ChEBI API and return results. Parsed the XML response using BeautifulSoup and structured the data for rendering in the template.
Test	Tested various search queries to ensure the API integration returns correct and expected results. Verified that error handling works correctly for failed API requests.
Learn	Documented the performance of the API calls, noting any latency or reliability issues. Considered user feedback on the types of searches and results displayed.

Design	Our goal was to create a selection mechanism for users to choose which ChEBI compounds to submit for further processing and to come up with the submission workflow to gather additional parameters including pH, temperature and aerobicity.
Build	Implemented checkbox functionality for selecting compounds in the search.html template. Utilize JavaScript to manage the selected compounds and send a structured JSON object to the Django backend upon form submission.
Test	Tested the compound selection and submission process with various scenarios (no selection, single selection, multiple selections). Validated the structure of the JSON sent to the server manually and ensured it matches the expected format.
Learn	Assessed user experiences with the selection process and submission workflow. Identified possible points of confusion for the user and crafted appropriate instructions to be integrated into the webpage to enable users of varying technical familiarity to use Astrolabe.

Design	Planned to draft the backend processing of selected compounds, utilizing parallel processing for efficiency and define the integration points for the other APIs (UniProt, BioCyc) within the processing workflow.
Build	Implemented the process_search_data function in views.py to handle data processing and API calls in parallel using multiprocessing. Ensured that each API function (e.g., kegg, uniprot, biocyc) can handle the input data properly.
Test	Validated that the multiprocessing implementation functions correctly and that each API call returns results. Tested performance improvements compared to sequential processing.
Learn	Analyzed the efficiency gains from parallel processing and document the overall execution time for processing requests. Reviewed error handling for API calls in a multiprocessing context.

Design	Developed a scoring mechanism to evaluate and rank the results from the integrated APIs based on user-defined parameters (pH, temperature). Planned the presentation of results to users in a clear and informative manner.
Build	Implemented the scoring_function in utilities.py to calculate scores based on combined results from the APIs. Created a results display section in the search.html template to show the top organisms based on the scoring results.
Test	Verify that the scoring function works correctly with various inputs and produces expected rankings. Ensure the results are displayed correctly and are easy to understand.
Learn	Collect user feedback on the usefulness of the presented results and the scoring mechanism. Analyze how well the results meet user expectations and make any necessary adjustments based on feedback.