Evaluation of deep eutectic solvent pretreatment towards efficacy of enzymatic saccharification using multivariate analysis techniques

Abstract

In this paper, deep eutectic solvent (DES) pretreatment of lignocellulosic biomass was analysed using principal component analysis (PCA) and partial least square (PLS) methods to determine correlations between the pretreatment process variables and assess their significance level. The analysis was conducted on the dataset containing 103 experimental observations and 32 variables. Both PCA and selected PLS model (M4) had substantial prediction power and prediction relevance, as exhibited by their good cumulative explained variance (R2cum, 0.787) and cumulative predicted variance (Q2cum, 0.524) for six principal components of PCA model, and explained variance of independent variable (R2Xcum, 0.781), explained variance of dependent variable (R2Ycum, 0.803) and cumulative predicted variance (Q2cum, 0.776) for the best PLS model. The results of PCA and PLS analyses revealed that the most significant variables for DES pretreatment were severity factor , temperature, type of antisolvent, particle size, agitation intensity and type of hydrogen bond donor (HBD) of the DES. The use of carboxylic acid and hydroxy acid HBD was associated with higher glucose yield and shorter duration of saccharification . The models also showed that enzymatic saccharification was more efficient when organic solvent or inorganic salt solution was used as antisolvent for lignocellulose regeneration. Usage of water as antisolvent was associated with reduced glucose yield during saccharification. Another notable outcome of this work was that intensive stirring during pretreatment had negative effect on glucose yield due to the shear thickening properties of DESs. This work provided a systematic analysis of DES pretreatment process of lignocellulosic biomass using PCA and PLS methods that allowed better understanding of the factors affecting the pretreatment process. • Multivariate analysis of PCA and PLS were used to study DES pretreatment. • The models had high predictive accuracy with good R2Xcum, R2Ycum, Q2cum values. • SF, antisolvent, particle size, agitation and HBD had great impact on pretreatment. • Long p-time, low t, AS (organic, inorganic), HBD (–COOH, –OH) favored pretreatment. • High agitation intensity and large particle size led to poor pretreatment.