METHOD FOR DETECTING CONTAMINATION IN MICROALGAE AND CYANOBACTERIA CULTURES

Abstract

The invention consists of a method for determining the purity of microalgae cultures, allowing the determination of the purity of the microalga of interest, as well as the percentage of contamination by other microalgae that may commonly grow alongside it. This method uses only a spectrophotometer to obtain absorbance scans between 350 and 750 nm. These scans are then normalized and analyzed using a 1D convolutional neural network trained to recognize the species of interest. The result of the invention is a quantification of the culture purity of the species of interest, as well as the percentage of other contaminants present. This allows for the rapid characterization of cultures, is low-cost, and is easily applicable in any microalgae production facility, including those for cyanobacteria.

Description

DESCRIPTION METHOD FOR DETECTING CONTAMINATION IN CROPS OF MICROALGAE and CYANOBACTERIA OBJECT OF THE INVENTION The object of the present invention is a method for detecting contamination in microalgae cultures by analysis using a neural network of spectrophotometric absorbance sweeps. The present invention falls within the field of microalgae production (including cyanobacteria) for any application, especially on an industrial scale. More specifically, the invention relates to determining the purity of cultures. The invention allows for determining the purity of cultures by confirming the desired microalgae species and the proportion of any other potential contaminants present. BACKGROUND OF THE INVENTION Microalgae production (including cyanobacteria) is a continuously growing industrial activity due to the wide range of applications these microorganisms offer. These include the production of food and nutraceuticals, cosmetics, animal and aquaculture feed, biostimulants and biopesticides for agricultural use, and even biofuels. All these applications require the stable and safe production of high-quality biomass, free from contaminants or unsuitable materials that could negatively impact their use. To carry out this production, different types of reactors, both open and closed, are used. These reactors are designed to control operating conditions so that the microalgae cells (including cyanobacteria) are in their most favorable state and grow with the highest possible productivity. These reactors are installed on large areas of land, on the order of hectares, and it is very common to install multiple units in the same facility since the maximum reactor size is limited by technological constraints. Regardless of the type of reactor used for microalgae production, the culture media are prepared by enriching water with nutrients so that the microalgae can photosynthesize, producing large quantities of oxygen by consuming large amounts of CO2, and transforming the nutrients into microalgae biomass.Specifically, these types of systems can produce up to 100 tons / ha·year of biomass, for which it is necessary to supply 200 tons / ha·year of CO2, as well as 10 tons / ha·year of nitrogen and 2 tons / ha·year of phosphorus, in the process also producing 150 tons / ha·year of O2. Microalgae production can be carried out in various systems, including closed and open photobioreactors. Closed photobioreactors allow for greater control of culture conditions, which can result in higher productivity and quality of the biomass produced, especially due to better control of potential contamination. On the other hand, open photobioreactors, such as raceway reactors, are more economical and suitable for large-scale production. However, in these systems, the risk of contamination is very high, both from other organisms and from other genera of microalgae that can colonize and compete with the genera of interest. This can affect the quality and productivity of the biomass. Therefore, it is essential to maintain control and monitoring of the culture to avoid this problem. Currently, this process is carried out by microscopic observation of the samples, which requires extensive experience and taxonomic knowledge, necessitating highly skilled workers (Chong et al., 2023). To facilitate this process, the use of equipment such as FlowCam for microalgae image acquisition and neural networks for classification has recently been proposed (Otálora et al., 2021). This has allowed for the replication of the laboratory process, but in a faster and more precise manner, although with the disadvantage that the use of FlowCam requires a high investment. Similar work has been carried out using a Specim IQ spectral imager and a one-dimensional convolutional neural network that allows for the determination of the concentration of five microalgae genera used for calibration (Salmi et al., 2022).Despite the adequate accuracy of these methods, both have the problem of requiring the use of expensive equipment by personnel skilled in its operation. SUMMARY OF THE INVENTION The invention describes a method for detecting contamination in microalgae and cyanobacteria cultures based on absorbance scan analysis comprising the following steps: - Perform absorbance scans on a culture using visible range spectrophotometry. - Calculate a trend line from the absorbance sweeps. - Subtract the trend line from the absorbance sweeps to obtain absorbance peaks, which contain the most relevant information for each genre. - Apply 1D convolutional layers of a neural network configured to extract patterns of values, shapes, and peaks in absorbance sweeps to the absorbance peaks. - Classify the absorbance peaks by relating their patterns and specific characteristics to a genus of microalgae or cyanobacteria using a Softmax activation function. - Use the Softmax activation function to obtain a probability vector that indicates the probability of each genre being the correct one, and based on this vector, the majority sample is indicated and whether it presents other minority genres that are considered as contaminants. The previously mentioned absorbance scans are performed between 350 and 750 nm and at intervals of 0.5 nm. Additionally, the method may also include a signal preprocessing stage using a baseline removal technique in which an algorithm is used to correct the baseline in spectral data, removing unwanted background trends. The method may also include an additional stage in which the sweeps and the results from the peak classification stage are converted into a user-readable format. This method allows the absorbance sweep stage to be carried out by manual sampling or by online microalgae culture characterization systems. Additionally, the method can be applied to samples from biomass production, wastewater treatment, or biofuel production. The method for detecting contamination in microalgae and cyanobacteria cultures according to the preceding claims, wherein the absorbance sweep stage is performed in open or closed reactors. The invention also relates to a system for detecting contamination in microalgae and cyanobacteria cultures based on absorbance scan analysis, comprising: - A spectrophotometer configured to perform absorbance scans on a crop in the visible range. - A processing module configured for: or calculate a trend line from the absorbance sweeps; or subtract the trend line from the absorbance sweeps to obtain absorbance peaks, which contain the most relevant information for each genre; or apply 1D convolutional layers of a neural network configured to extract specific patterns and features to the absorbance peaks; and or classify absorbance peaks by relating their specific patterns and characteristics to a genus of microalgae or cyanobacteria using a Softmax activation function. BRIEF DESCRIPTION OF THE DRAWINGS To complement the description being made and to aid in a better understanding of the characteristics of the invention, according to a preferred embodiment thereof, a set of drawings is attached as an integral part of said description, in which, for illustrative and non-limiting purposes, the following has been represented: Figure 1 shows an actual image of a Scenedesmus culture. Figure 2 shows two spectrophotometry graphs in which wavelength is displayed versus absorbance, the upper one corresponding to a scan in the visible range, and the lower one to the normalized scan, which allows quantification of the purity and / or presence of other microalgae by means of the developed neural network. PREFERRED EMBODIMENTS OF THE INVENTION The present invention discloses a method for detecting contamination in microalgae cultures (including cyanobacteria) that is applicable to any microalgae production process and allows for quick and easy confirmation of the purity of the microalgae of interest, as well as the possible presence of other contaminating microalgae. The type of cultivation or application of microalgal biomass (including cyanobacteria) to which the present invention can be applied is not critical. It can be used for the production of biomass for human and / or animal consumption, as well as for any other application such as wastewater treatment or the production of biomaterials, including biofuels. Furthermore, the application is usable in any type of reactor, whether open, including thin-layer raceway reactors, or closed, such as bubble columns or tubular reactors. The invention facilitates decision-making regarding preventive and corrective measures for pollution control, as well as the use of biomass for various applications where purity is more or less relevant. An optical spectrometer was used to obtain an absorbance scan in the 350-750 nm range. This type of instrument is commonly used for nutrient analysis in crops, and is therefore frequently found in microalgae production facilities. The method involved developing a neural network based on 1D convolutional layers and following deep learning concepts to process absorbance data and classify different genera of microalgae. 1D convolutional layers are particularly well-suited for working with sequential data, as in this case, where each sample is a spectrum. The neural network was trained using a pre-labeled dataset, which included spectra from different genera of microalgae. During training, the network learned to identify specific patterns and features in the absorbance spectra of each sample, such as peak patterns. This allowed it to establish non-linear relationships between the different input signals, or absorbance spectra, and the different genera to be classified. Once trained, it was able to classify new microalgae samples based on their spectra. The resulting model was trained with laboratory samples but was subsequently validated with samples from industrial-scale photobioreactors operating under different conditions, both indoors and outdoors.The results showed high accuracy and efficiency in all tests performed, demonstrating this solution as a suitable methodology for detecting contamination in microalgae cultures. The results were validated using digital microscopic images as a reference to confirm the results of the proposed model. Thus, the proposed methodology consists of several parts for perfect operation: i) Visible range spectrophotometer that allows absorbance sweeps between 350 and 750 nm at intervals of 0.5 nm. ii) Sweep normalization process, which allows normalizing the activations in a layer by calculating the mean and variance of each sweep. The signals are normalized to have a mean of zero and a variance of one. This stabilizes the training, accelerates convergence, and reduces the risk of overfitting. iii) Application of 1D convolutional neural network for processing the normalized sweeps, identifying the different species and generating the final results of the model. iv) Handling of the sweeps and results for their conversion into the final reading of the purity of the microalgae of interest and estimated degree of contamination by other species. In one particular embodiment of the invention, it is used to verify that the Spirulina biomass produced in a reactor is free of contaminants from other microalgae. Culture samples are taken, and the absorption sweep between 350 and 750 nm is determined. This sweep is fed into the neural network, and by comparison with the sweeps previously used to train the network, the network determines that the culture is 99.98% free of other microalgae. In another test performed with a partially contaminated reactor, the same procedure is followed, and it is concluded that the culture is 5% contaminated by Scenedesmus. Microscopic observation of the culture corroborated this result, demonstrating the effectiveness of the developed model. In a preferred embodiment, a signal preprocessing stage is described using a baseline removal technique in which an algorithm is used to correct the baseline in spectral data, removing unwanted background trends. Additionally, the method includes a stage for estimating contamination by other microalgae using a Softmax activation function in the last layer of the network. This layer generates a probability vector as output, indicating the probability of each genus being the correct one. A key characteristic of Softmax is that the sum of the probabilities for all classes (genera) must always equal 100%. In the case of pure samples, the network typically assigns probability values ​​greater than 96% to the analyzed genus. Therefore, by evaluating the values ​​of this vector, it is possible to determine whether the analyzed culture is contaminated based on the value of the dominant genus. In comparison with what is described in the state of the art, this work has addressed a simpler and more economical approach for the classification of microalgae genera and the determination of both the purity of the cultures and the percentage of contamination by other genera. The main advantages derived from the procedure of the present invention are: - It is a low-cost methodology, requiring no additional equipment beyond that usually available in any microalgae production facility. - It is easy to use as it does not require extensive experience in taxonomy or handling sophisticated equipment. - It is fast because it allows identification of the microalgae present in the culture in little more time than it takes to obtain the absorbance scan of the sample, in any case less than 5 minutes - It is compatible with other technologies such as digital image processing or evaluation by omics methods, which allow the results to be validated but involve more time and higher cost. - It is applicable to any microalgae provided that the network is properly trained with samples of said microalgae. The novelty of the system lies in its ability to provide rapid and precise measurement of the quality of biomass produced in a microalgae photobioreactor. Quality is defined as the degree of purity of the microalgae being produced, as well as any potential contamination. The proposed invention is of great interest for any application derived from the use of microalgae, although it is particularly valuable for the production of pure strains for high-value applications such as pharmaceuticals, cosmetics, nutraceuticals, and human food.

Claims

1. A method for detecting contamination in microalgae and cyanobacteria cultures based on absorbance sweep analysis comprising the steps of: - performing absorbance sweeps on a culture using visible-range spectrophotometry; - calculating a trend line from the absorbance sweeps; - subtracting the trend line from the absorbance sweeps to obtain absorbance peaks, which contain the most relevant information for each genus; - applying 1D convolutional layers of a neural network configured to extract patterns of values ​​to the absorbance peaks,1. The method for detecting contaminants in microalgae and cyanobacteria cultures according to claim 1, wherein the absorbance scans are performed between 350 and 750 nm and at 0.5 nm intervals.

2. The method for detecting contaminants in microalgae and cyanobacteria cultures according to the preceding claims, wherein the absorbance scans are performed between 350 and 750 nm and at 0.5 nm intervals.

3. The method for detecting contaminants in microalgae and cyanobacteria cultures according to the preceding claims,further comprising a signal preprocessing step using a baseline removal technique in which an algorithm is used to correct the baseline in spectral data, eliminating unwanted background trends.

4. The method for detecting contamination in microalgae and cyanobacteria cultures according to the preceding claims, further comprising a step of converting the scans and the results of the peak classification step into a user-readable format.

5. The method for detecting contamination in microalgae and cyanobacteria cultures according to the preceding claims, wherein the absorbance scan step is carried out by manual sampling or by online microalgae culture characterization systems.

6. The method for detecting contamination in microalgae and cyanobacteria cultures according to the preceding claims,wherein the absorbance sweep stage is performed by taking samples from biomass production, wastewater treatment, or biofuel production.

7. The method for detecting contamination in microalgae and cyanobacteria cultures according to the preceding claims, wherein the absorbance sweep stage is performed in open or closed reactors.

8. A contamination detection system in microalgae and cyanobacteria cultures based on absorbance sweep analysis comprising: - a spectrophotometer configured to perform absorbance sweeps on a culture in the visible range; - a processing module configured to: perform absorbance sweeps on a culture using visible-range spectrophotometry; calculate a trend line from the absorbance sweeps; subtract the trend line from the absorbance sweeps to obtain absorbance peaks,containing the most relevant information for each genus; applying 1D convolutional layers of a neural network configured to extract patterns of values, shapes, and peaks in the absorbance sweeps to the absorbance peaks; classifying the absorbance peaks by relating their patterns and specific characteristics to a genus of microalgae or ianobacteria using a Softmax activation function; and using the Softmax activation function to obtain a probability vector indicating the probability of each genus being the correct one, and based on this vector, indicating the majority sample and whether it contains other minor genera that are considered contaminants.

Images