Nitrate and nitrite as mixed source of nitrogen for Chlorella vulgaris: fast nitrogen quantification using spectrophotometer and machine learning

This article presents a machine learning workflow allowing to construct spectrophotometric equations predicting nitrate and nitrite concentrations within microalgae culture samples. First, numerous samples with various nitrate and nitrite concentrations (in mixture or separated) were drawn from cultures. Their UV absorbance spectra were recorded with a tabletop spectrophotometer before being analyzed using ion chromatography. Then, the data collected were used to construct a machine leaning model based on partial least square regression. From a practical perspective, the best model involves 3 wavelengths to quantify both nitrate and nitrite within the samples. The proposed equations can readily be used (LoQ of 0.5 mg L- 1, uncertainty of +/- 10%). They greatly shorten the time to obtain sample nitrate and nitrite concentrations compared to ion chromatography while retaining adequate accuracy. Furthermore, the workflow is presented step-wise, with emphasis on relevant details so that other scholars may deploy in their own laboratory to best suit their own needs. Finally, the data and source files are made available in an online repository.

Automated summary

The article introduces a machine learning workflow to construct spectrophotometric equations predicting nitrate and nitrite concentrations in microalgae culture samples. The workflow involves recording UV absorbance spectra of samples, constructing a machine learning model based on partial least square regression, and utilizing 3 wavelengths to quantify nitrate and nitrite concentrations. The proposed equations provide a faster and more accurate alternative to ion chromatography for determining sample concentrations.