Machine learning-based screening for ionic liquid pretreatment of plant biomass for biofuel production

Lignin, the second most prevalent biopolymer in nature, has garnered attention as a promising reservoir for sustainable fuels, chemicals, and materials. Overcoming significant hurdles, such as identifying suitable solvents and implementing cost-effective and efficient processes for lignin dissolution and depolymerization, stands as a pivotal challenge in the quest to transform lignin into value-added products. While some ionic liquids (ILs) exhibit the capacity to dissolve and depolymerize lignin, crafting an effective IL specifically tailored for lignin dissolution remains a formidable task. To address this issue, the COnductor-like Screening MOdel for Real Solvents (COSMO-RS) model was used to screen ILs by computing logarithmic activity coefficients (ln(γ)) of lignin. Based on these simulations, we develop a machine learning model that can accurately predict the activity coefficients only using properties computed from the ILs structure. Furthermore, we develop a pipeline that integrate overall biomass compositions and using hierarchical clustering identify ILs that tailored to that particular biomass composition.