Chemometric modeling of UV-visible and LC-UV data for prediction of hydrolysate fermentability and identification of inhibitory degradation products.
Access rightsWorldwide access.
Access changed 5/21/14.
MetadataShow full item record
Production of ethanol from lignocellulosic biomass requires a pretreatment step to liberate fermentable sugars trapped within the plant. During pretreatment, lignin and some sugars undergo degradation to form compounds which have shown inhibitory effects to fermentative microorganisms. Accordingly, development of a rapid and accurate method for assessment of microbial inhibition and identification of inhibitory compounds is essential for gaining a better understanding of pretreatment and its downstream effects on fermentation processes. Traditional methods for identification of inhibitory compounds involve a “bottom-up” approach. Using this approach, one or more known degradation compounds are added to fermentation media and their effects on batch fermentation of ethanol are observed. These methods are extremely time-consuming and labor-intensive which makes them unattractive to researchers. Furthermore, they are carried out on degradation compounds that have already been identified. Given that biomass hydrolysates contain many unidentified constituents, identification of inhibitory compounds by traditional means is unlikely to occur on a timescale that is consistent with current mandates for commercial production of cellulosic ethanol. To address these limitations, we have developed a chemometric model that correlates ultraviolet (UV)-visible spectroscopic data of 21 different biomass hydrolysates with their fermentability (percent inhibition of ethanol production). This novel approach enables rapid prediction of hydrolysate fermentability using UV-visible spectroscopic data alone and offers significant improvements in throughput and labor when compared to traditional batch fermentation methods. The model was subsequently used to predict percent inhibition for five hydrolysate samples, with a root-mean-square error of prediction of 6%. To evaluate the use of chemometric modeling for identification of inhibitory compounds in biomass hydrolysate, a second model was developed to correlate HPLC-UV chromatographic data of the 21 hydrolysates with their percent inhibition. Detection was monitored at four specific wavelengths identified by the UV-visible model as significant spectral regions. Once constructed, the HPLC-UV model was used to identify retention times that had the highest correlation with inhibition. To determine whether better resolution or more universal detectability of sample constituents may lead to identification of additional retention times, a third chemometric model was developed with chromatographic data of hydrolysates obtained via ion chromatography with conductivity detection.