Latest revision as of 10:59, 18 October 2019

Nit_Yellow

A full description of the part: This composite part is a nitrate biosensor. It is composed of the nitrate sensitive promoter PyeaR, a strong rbs, the amilGFP yellow chromoprotein and a transcriptional terminator. In presence of nitrate, bacteria turn into different yellow color intensities according to the concentration of nitrate.

UPNAvarra_Spain 2019

BBa_K2817007 was first designed by Zhaoyu Liu from team NEU_China_A in iGEM 2018. It is a nitrate reporter, PyeaR-amilCP composite. In order to improve this system, we replaced the amilCP with the amilGFP protein. Although both enzymes can be visualized by human eyes, so both are suitable to our purpose, it was previously reported (iGEM Bulgaria 2018 team) that amilGFP is more stable upon re-cultivation, so it could be a better reporter candidate.

This modified biosensor is composed of the nitrate sensitive promoter PyeaR, a strong RBS, the amilGFP yellow chromoprotein and a transcriptional terminator. In presence of nitrate, bacteria turn into different yellow color intensities according to the concentration of nitrate. The correct construction of this plasmid was confirmed by sequencing (Figure 1).

Figure 1. Construction of expression vector Nit_Yellow (BBa_3287001) from parts of 2009_Edinburgh: BBa_K216005; and amilGFP coming from the chromoprotein collection of 2011_Uppsala-Sweden team. The PyeaR-amilGFP composite part is cloned in the pSB1C3 vector through the BioBrick suffix-prefix site.

We transformed the plasmid containing BBa_K2817001 into E. coli competent cells and cultured at 37ºC until OD = 0.4. Then we add KNO₃ al different concentrations to induce the expression at 37 ℃ for 6 hours. After that, 2 mL of the bacterial culture were centrifuged at 3,000 r.p.m for 3 minutes, so we could observe at first sight the result of PyeaR promoter being activated by nitrate (Figure 2).

Figure 2. AmilGFP expression levels under increasing concentrations of KNO₃.

For these experimental results we generated a mathematical model, in order to prove that the (imaging) data we have gathered in the lab is in fact learnable by a simple regression model. We have opted out by a standard Least-Square error (linear) regression model, which has been run on the dataset obtained in the imaging part. This dataset consists of the average RGB color in the colored part of the pellets used at different concentrations of KNO₃ (Figure 3A). For each color, we have subselected the channels that we are interest for the problem. That is the Red channel in this case. It can be seen how the data is easily learnable by a linear regression model (Figure 3B) and, moreover, the error in the model training is small (0.17).

Figure 3. Modeling a Nitrate biosensor. A) Input data; B) Regression model.

If we compare these results with the construction harbouring the amilCP chromoprotein (BBa_K2817007 or our twin BBa_K3287000), we can see that both are gradually sensitive to the nitrate concentration in water, but the expression of the amilGFP chromoprotein is more linear over the best visible channel. In this sense, it looks like a better candidate for visual detection and quantification of the presence of nitrate in water. Although the difference is minimal (Avg.Error_Blue=0.18 vs Avg.Error_Yellow=0.17), specially given the short number of experiments, it has been sustained over several lab repetitions.

We have proved that data can be learnable by a simple linear regression model (OLS), but to obtain more accurate predicted values, we trained our model using more data to be more precise using machine learning techniques. These extra data are extracted from images taken from pellets induced by a known concentration of KNO₃. In order to test our model, we have applied a cross validation. We divided our data in n partitions, each one with 25% of the data. Our training data will be n-1 partitions whereas one will be our test data. Every partition is selected once to be the test data so the training and testing is done 4 times.

Figure 4. Modeling a Nitrate biosensor using machine learning.

If we compare again these results with the construction harbouring the amilCP chromoprotein (BBa_K2817007 or our twin BBa_K3287000), we can see that both are gradually sensitive to the nitrate concentration in water, but the amilGFP chromoprotein gives us a minor average error (Avg.Error_Blue=0.289>Avg.Error_Yellow=0.236), and is more linear over the best visible channel (R²_Blue=0.930<R²_Yellow=0.949). So again, in this sense, it looks like a better candidate for visual detection and quantification of the presence of nitrate in water.

Applying a dimensionality reduction in order to reduce the computational complexity of our model using PCA (Principal Components Analysis), and comparing once again these results with the construction harbouring the amilCP chromoprotein (BBa_K2817007 or our twin BBa_K3287000), the amilGFP chromoprotein gives us a significant minor average error (Avg.Error_Blue=0.400>Avg.Error_Yellow=0.232) and is more linear over the best visible channel (R²_Blue=0.838<R²_Yellow=0.950). So once again, it looks like a better candidate for visual detection and quantification of the presence of nitrate in water.

Figure 5. Modeling a Nitrate biosensor using PCA.

Sequence and Features