Comparative assessment of two bioproducts using quantitative image analysis

Polyhydroxyalkanoates (PHA) and triacylglycerides (TAG) could be used as bioplastic and precursors of biodiesel, respectively. To ensure a more controlled and efficient operation of the reactor, rapid monitoring of bioproducts accumulation is essential for enabling precise process adjustments and efficiency. Usually, gas chromatography (GC) or high-performance liquid chromatography (HPLC) are the methods used for bioproduct quantification, but these are hazardous and labor-intensive techniques. In this context, quantitative image analysis (QIA) emerges as a rapid and non-hazardous alternative for bioproducts quantification. The present work investigates the application of QIA for the simultaneous quantification of PHA and TAG in a mixed microbial culture fed with used cooking oil (UCO). Three stains were evaluated: Sudan Black B (SBB), Nile Blue A (NBA), and BODIPY. The data obtained from the images acquired were processed using multivariate statistical methods, specifically Partial Least Squares (PLS) and Multilinear Regression (MLR). Overall, SBB showed comparatively better results for PHA prediction, and the PLS models generally demonstrated higher reliability than MLR. However, none of the tested stains were effective in quantifying TAG, likely due to the concentration of data around a limited range, which hindered the model´s ability to predict this bioproduct. Simultaneous quantification of both products (PHA+TAG) resulted in intermediate model performance (R 2 ≈ 0.62-0.89), without improving the predictive accuracy of PHA quantification alone. These findings support the potential of image-based techniques for intracellular bioproduct quantification, while also highlighting the need for a larger and more diverse dataset to enhance model prediction accuracy and robustness. • Quantitative image analysis was used for bioproducts quantification. • Sudan Black B showed better results for polyhydroxyalkanoates quantification. • Partial least squares generally showed better results than multilinear regression. • Challenges in predicting triacylglycerides arise with datasets of low variability.