Cobalt chloride was one toxin type that the algae cells were inoculated with. Cobalt is a heavy metal that enters waterways through the weathering of soil and rock formations. Therefore, it is a toxin which may be found near industrial mining operations. This adds an additional stakeholder: industrial mining companies that must monitor cobalt contributions to waterways to satisfy legal regulations (ATSDR 2022). 

Figure 2. displays one replicate of spectra that resulted after scanning pelleted algae cells that were inoculated with varying concentrations (0-250mM) of cobalt chloride for 60 minutes. There are some visible changes between the spectra of different samples, but it is difficult to interpret if the differences are due to toxin response or other factors such as resolution or pellet size. Two other experiments ran included inoculation of algae with the same concentration values, but for 6 hours and 24 hours.

The same concentration range was tested for ammonium sulfate as was tested for cobalt chloride. The experiments also included scans for 6-hour and 24-hour inoculation periods. The spectra in Figure. 2 and Figure 3. both include the normalized wavelength intensities, as produced by a normalization code developed in MATLAB by the Senger Lab. This technique is applied to remove the overbearing effect of fluorescent signal on interpretation of the Raman signal. An overbearing fluorescent signal occurs when a molecule contains wavelengths matching those of the laser. 

PCA is a chemometric modelling tool which detects areas of large variance in spectra between samples and groups them accordingly. Typically PC1 is the variable representing spectral range with the most dramatic difference, with PC2 and PC3 following. The PCA plot shown in Figure 4. highlights two areas of clustering representing two distinct phenotypes picked up by Raman spectroscopy in the same experiment depicted by the spectra in figure #. 

Unfortunately, the spectra from the other experiments were ran through a PCA model and did not exhibit the same degree of clustering. Therefore, the group decided to run the spectra and PCA through more advanced chemometric models

The goal of this modelling technique was to see if samples could be separated by treatment group. Figure 5. shows the MANOVA cluster analysis developed from all spectra taken from algae inoculated with a >1mM concentration of toxin- either ammonium sulfate, cobalt chloride, or copper sulfate. Although overlap was observed in the center, significant clustering among treatment groups was observed. 

Additionally, there is an interesting trend visible as you move from the left to the right side of the graph that may correlate with increasing toxicity of the treatments to the algae. Ammonium sulfate promotes cell growth, copper sulfate is slightly toxic to the cells, and cobalt chloride is the most toxic to the cells. It is important to note that less scans were performed on copper sulfate inoculated algae than the other two treatment groups.

We compared the commercial costs of some individual chemical detection tests that can be done on water samples to the cost of running a test using our solution, which will scan for all toxins in our spectral library. The initial cost for the handheld spectrometer and 3D printed attachment would be about $2010, but each additional scan would cost only about $12, with a majority of that price stemming from the cost of the algae culture- which can be easily grown by a consumer who is using this solution consistantly.