Research Projects

The following is an incomplete list of some research projects underway in the Burgin Lab.

Circular Plastic Bioeconomy

A diagram depicting the workflow for co-optimizing two steps of the PETase reaction with a neural network trained on simulations of each step. The model uses active learning to guide the collecting of more training data, eventually producing an optimized PETase, shown breaking down the PET in a water bottle into individual components.

Plastic recycling is essential in both reducing the production of plastic waste and in reducing reliance on fossil fuels used as raw materials. However, plastics are notoriously difficult to recycle efficiently in part due to their wide variety of chemical makeups, properties, and additives. In addition, imperfections in recycling processes result in degradation of physical properties of recycled plastics, limiting the number of times a given material can be efficiently recycled. These limitations motivate the development of alternative plastic recycling technologies with improved efficiency and repeatability.

A recently discovered enzyme, PETase catalyzes the degradation of polyethylene terephthalate (PET). Enzymes capable of specific decomposition of plastics into monomers under mild conditions are an ideal solution to plastic recycling, as monomers can be reused to produce new plastics without the degradation in material properties associated with mechanical or pyrolytic recycling strategies. However, despite recent efforts to improve the catalytic efficiency of PETase, the best available variants still require improvement in order to better compete with traditional recycling technologies. This project will leverage molecular models of both steps of the catalytic mechanism of PETase to train a machine learning model that co-optimizes both steps together and directs simulations to obtain additional training data through active learning.

Enzyme Engineering for Carbon Capture

Roughly half of all U.S. carbon dioxide emissions are attributable to fossil fuels burned during industrial manufacturing and electricity production. Post-combustion carbon capture at the effluents of major emissions sources offers an essential tool for meeting urgent emissions reduction goals and mitigating climate catastrophe in the near term while alternatives to fossil fuels are further developed. Unfortunately, available technologies for addressing this need are expensive and inefficient: capture with organic solvents presents issues with corrosion and regeneration after capture, and membranes suffer from fouling and inconsistent performance under flue gas conditions.

One of the most promising alternative options is carbonic anhydrase (CA), an exceptionally efficient type of metalloenzyme that converts carbon dioxide and water to bicarbonate and hydrogen. Although CA vastly outperforms alternatives in terms of efficiency, carbon footprint, ease of captured carbon recovery, and regeneration after capture, it suffers from instability over long timescales in the presence of harsh conditions in flue gas columns, due to high temperature and salinity, alkaline pH, and other issues. Engineering of CA variants that remain highly stable and catalytically active under these conditions will open up a new avenue for efficient and cheap carbon capture. This project will use a machine learning model trained on molecular simulations to predict stabilizing mutations and identify regions of the protein to focus on.

A diagram of a gas chamber containing carbonic anhydrase. Flue gas goes into the chamber at the bottom, and inside the chamber carbon dioxide and water is converted to bicarbonate and protons. Scrubbed flue gas exits the chamber at top.

Oligosaccharide Biomanufacturing

A two-panel figure showing, at left, a zoomed in diagram of a restrained reaction transition state in the active site of an enzyme, and at right, a scatter plot showing an inverse linear relationship between simulation transition state complex root mean square fluctation and experimental activity measurements across six different enzyme variants.

Polysaccharides are an essential class of functional biopolymers with wide-ranging roles, from structural support to signaling cascades and mediation of cell-cell interactions. In particular, short sugar chains known as oligosaccharides are essential in cell signaling and post-translational modification of proteins, and have applications as functional food additives with properties ranging from treatment of cancer and inflammation to promotion of gut microflora development in infants. Unfortunately, techniques for synthesizing oligosaccharides have lagged significantly behind those for other biopolymers.

Glycosynthases are engineered enzymes that present an attractive alternative to existing oligosaccharide synthesis pathways; however, these enzymes are too inefficient for industrial applications without further engineering. This project focuses on applying a novel molecular simulations approach to obtain training data for a machine learning model in order to accurately predict the effects of mutations on enzyme efficiency, helping to obtain improved variants.