In order to demonstrate that our bacterial biosensors can be used to detect and quantify pollutants in the ambient, we tested some water samples collected in differences places along Navarre, to prove our system applicability in the real life.
Firstly, we contacted with a cooperative, leader of agro-food sector in Spain with the aim of analyze water samples coming from different agro-industrial processes from different facilities of the cooperative in our region. We took samples from an avant-garde poultry processing center, a fourth and fifth range factory that produces bagged salads and also from the greenhouse where the vegetables used in this factory are grown. With this experience, our team not only got water samples to test our biosensors, but also we learnt many thing about the management of water in different agro-industrial processes and farming, and how they analyze its quality.
Given that nitrate accumulation in water is an environmental problem that directly concerns the region of Navarre, we took water samples from different rivers along the areas potentially affected by this problem. Following the EU Directive 91/676/CEE, addressed to the protection of waters from cross contamination from agricultural activities, different regions were detected (vulnerable zoned) and put under supervision for their potential health-damaging situations. Such regions are the hydrological basins of the rivers Cidacos, Robo and Ebro-Aragon. Our local government did not only restrict to comply with EU regulations, also passing regional laws (Orden Foral, in their local naming) 247/2018. This law regulates the control of the areas, including registered supervision and, additionally, established an action plan to be executed in the period 2018-2022. In the following map we indicated the specific locations from which samples were taken (Figure 1).
Figure 1. Areas potentially affected by nitrate accumulation and specific locations from which samples were taken.
Moreover, we can add the collaboration of two teams participating in this year competition. We received water samples from Athens and Costa Rica, ready to check their content of nitrate and heavy metals.
The aim of our project was to modify bacteria to make them change their colour if they were exposed to the presence of heavy metal or nitrates. However, we decided to take that idea to the next level. We thought that we could create an informatics application, which using a picture of the bacteria, could tell an approximately concentration of the substance that is being tested, depending on its colour.
We used image processing techniques, combined with machine learning algorithms, in order to create a model to predict the concentration of each specific substance (heavy metal or nitrate) from an image: just taking a photo of the colored pellet of the sample, the program is able to calculate the approximate concentration of the heavy metal or nitrate.
For each sample different inductions were performed in order to detect the presence of heavy metals or nitrate. After the induction, the pellets didn’t present a very bright tonality, so the correlation between concentration and colour couldn’t be appreciated to the naked eye (Figures 2-3).
Nevertheless, applying our machine learning model to the images from the colored pellets, we could detect the presence of heavy metals or nitrate, and even predict the concentration, due to its ability to distinguish small color variations. We applied this algorithms to all the samples we took from different industries and diverse rivers along the areas potentially affected by nitrate accumulation.
Figure 2. Sample with concentration of Nitrate close to 0.
Figure 3. Sample with high concentration of Nitrate.
As we expected, using our machine learning model, the nitrate concentration from the samples we took from the avant-garde poultry processing center, is close to 0 due to the frecuently mandatory water quality controls. The nitrate concentration from the samples we took from the fourth and fifth range factory, also as we expected, is close to 0. Other water samples we took from different industries, had a null concentration of nitrates.
We also analyze the water samples from the different rivers along the areas potentially affected by nitrate accumulation. On the Robo´s basin we took a sample that throws a nitrate concentration of 0.52 with our yellow chromoprotein. This is a representative example of how the correlation between concentration and color can not be appreciated to the naked eye, but with our model we can detect the presence of nitrate.
On the Cidaco´s basin, the sample throws a nitrate concentration of 0.4, as we expected, so we can assure that there is a high risk of nitrate contamination. In the case of the Azgagra Ebro’s basin, we obtained a concentration of 0.35. Every water sample from the areas potentially affected by nitrate accumulation, presented a high nitrate concentration. Hence, we can again assure that there is a high risk of nitrate contamination in this areas of southern Navarre.
We also measured the mercury and cooper concentrations of the samples from the vulnerable zones rivers. As we expected, we detected a very low concentration of mercury and cooper close to zero: a concentration lower than the one established by the Environmental Protection Agency (EPA). Hence, we can affirm that these concentrations had no risk for human health or the environment.
Thanks to this experience, we successfully proved that our project has a real application and we demonstrated its utility.
