Branco Weiss Fellow Since
2020
Research Category
Glycobiology, Machine Learning
Research Location
Department of Chemistry and Molecular Biology, and Wallenberg Centre for Molecular and Translational Medicine, University of Gothenburg, Sweden
Background
As humans, we are covered with sugar. Not only our skin but every surface imaginable – organs, cells, lipids, proteins – is coated with long chains of carbohydrates or glycans. This sweet shell is in fact present on any biological entity, from animals over plants to microbes and even viruses. Yet while we are aware of DNA as our heritage and of proteins as the movers and shakers in our cells, glycans remain curiously hidden in plain sight. Few and far between are the textbooks or popular science illustrations that depict a cell or protein with its associated glycans. Despite being neglected by most scientists, glycans are the ultimate modulators, uniquely influencing the functions of nearly all proteins, cells, and organisms – in both health and disease – and are crucial for host-pathogen interactions. Therefore, those few researchers willing to wade into glycobiology have long sought to leverage their abundant potential for understanding biological processes as well as to use them as new kinds of diagnostics and therapeutics.
Yet working with glycans entails challenges. As the only nonlinear biological sequence, glycans can be branched. They are also the only biological sequence that does not have a largely universal alphabet across species (as, for instance, the A,T,G,C of DNA) and exhibit an astounding diversity of different building blocks as well as ways to connect them. All this has stymied progress in understanding this layer of complexity in biology. Connecting sequences with functions always represents a central goal in the analysis of biological sequences, as it facilitates mechanistic understanding – and how to disrupt those mechanisms. Past research has come far in enabling this feat for DNA, RNA, and proteins via predictive models based on machine learning, network analyses, and bioinformatics. Yet this type of analysis is still largely lacking for glycans.
Details of Research
In his research, Dr. Bojar is working to overcome this obstacle and transform glycobiology into a discipline that is predictable, unlocking it for biomedical applications in the process. Dr. Bojar was the first researcher to devise, develop, and apply methods derived from deep learning and natural language processing to glycans, learning the rules and grammar of the “third language of life,” next to DNA and proteins. Taming the incredible complexity of glycans, these models now can extract functions from glycan sequences and aid in understanding the overarching roles of glycans in biology. Supported by the Branco Weiss Fellowship, Dr. Bojar now engages in the ambitious task of translating these methods to the biomedical realm by developing glycan-based antibiotics and antivirals.
As glycans can block viral cell entry and microbial cell adhesion, they represent a treasure trove of novel potential drugs. Constructing machine learning models to predict antibiotic and antiviral activity from glycan sequences will thus uniquely enable Dr. Bojar to leverage the diversity of glycans to design candidate glycan therapeutics. These glycans can then be synthesized, chemically or enzymatically, and tested for their activity to curb microbial growth or viral infectivity. Any information will be fed back to the models, facilitating a design-build-test cycle that will lead to a continuous improvement of the candidate glycan therapeutics. In the age of antibiotic resistance and global pandemics, novel solutions to these vexing problems are needed. With his work, Dr. Bojar envisions giving new impetus to drug development in these areas but also to our understanding of glycans in biological processes in general, and thereby to the understanding of ourselves as humans.
