Systematic Review of Machine-learning Techniques to Support Development of Lignocellulose Biorefineries

Abstract

Lignocellulosic biorefineries (LBRs) are platforms for the production of a variety of bio-based products such as biofuels, biomaterials, biochemicals, food, and feed using lignocellulosic biomass (LB) as feedstock. LBRs are still rare worldwide. Their commercialization depends on challenges associated with the entire feedstock supply chain, efficiency, sustainability, and scale-up of pretreatment methods, as well as isolation and purification of value-added products. Each step within LBRs requires the development of new technologies or the improvement of existing ones, considering all three sustainability dimensions, environmental, social, and economic. Machine learning (ML) methods are widely used in various industrial fields, including biotechnology. The merging of biotechnology and ML has driven scientific progress and opened new opportunities for the development of LBRs as well. In this review, ML methods and their efficiency, used in biotechnology (metabolic engineering, bioprocess development, and environmental engineering), are presented, followed by their application in various phases of LB valorization.