Where code meets research: Collaboration at the heart of innovation

Carbohydrates are among biology’s most structurally diverse molecules. Those on bacterial surfaces play important roles in infection and can be recognised by the immune system, making them valuable vaccine targets. They often contain unusual sugars, branching patterns, repeating units and chemical modifications, while their flexible three-dimensional structures influence how they interact with antibodies. 

Professor Michelle Kuttel, based in UCT's Department of Computer Science, and her collaborators have spent more than 15 years developing an interconnected software ecosystem that helps researchers build, store, visualise and interpret complex carbohydrate structures. Together, these tools support computational glycoscience and research into bacterial vaccines.

"For carbohydrates, the challenge is predicting three-dimensional shape from molecular structure and then relating that to the complex immune responses to these diverse molecules," Kuttel said. 

Addressing this challenge has been made possible by a collaboration of more than 15 years between Kuttel and Professor Neil Ravenscroft of UCT’s Department of Chemistry that combines her expertise in computational glycochemistry, software development and visualisation with his knowledge of carbohydrate chemistry, which is crucial to the scientific validity of the software. Dr Nicole Richardson, a postdoctoral fellow in the Vaccine Group, links molecular modelling, and software development with experimental carbohydrate chemistry and vaccine-focused applications, while successive generations of Honours, Master's and PhD students have expanded and refined the group's tools.

An early challenge was how to visualise carbohydrates. Traditional atom-and-bond representations can obscure important structural features such as branching, backbone organisation and unusual ring conformations. To address this, the team developed specialised visualisations, including  PaperChain and Twister, in collaboration with John E. Stone, one of the principal developers of the Visual Molecular Dynamics software. This work evolved into CarbVis, an open-source extension for the UCSF ChimeraX molecular visualisation platform. CarbVis was developed by Honours student André Toerien, with research input and evaluation from Nicholas Yerolemou, Dr Nicole Richardson and Carla Coetzee. CarbVis highlights the structural features carbohydrate researchers need to interpret complex molecules and simulations. Its development reflects a principle that has guided the group's work from the outset: scientific software should be designed in close collaboration with the researchers who use it.

The team's next challenge was helping researchers construct scientifically plausible three-dimensional carbohydrate models for simulation. Building these structures manually is time-consuming and requires specialist knowledge, particularly for the unusual carbohydrates found in bacteria. To streamline this process, they developed CarbBuilder: software that converts a concise textual description of a carbohydrate into a three-dimensional molecular model. The tool incorporates knowledge of carbohydrate nomenclature, stereochemistry and molecular geometry, allowing researchers to generate reliable models more efficiently and consistently.

Originally developed through an Honours project, CarbBuilder was expanded through collaboration with Professor Göran Widmalm of Stockholm University. It has supported research on carbohydrate antigens from pathogens including Streptococcus pneumoniae, Group B Streptococcus, Shigella flexneri and Haemophilus influenzae. Dr Richardson has been instrumental in connecting these models and simulations with experimental chemistry and vaccine research.

The SugarDB web-based database allows researchers to search and explore these structures, automatically generates standard glycan diagrams and provides specialised carbohydrate-research tools. 

The SugarDB webtool addresses another challenge: bacterial carbohydrate structures are scattered across publications and difficult to compare or reuse. The web-based database allows researchers to search and explore these structures, automatically generates standard glycan diagrams and provides specialised carbohydrate-research tools. Ph.D student Carla Coetzee is currently expanding its structural data and refining its interface. One focus is Klebsiella pneumoniae, a major cause of neonatal sepsis and an antimicrobial-resistance threat. SugarDB aims to help researchers identify shared structural features that may contribute to antibody cross-reactivity and inform vaccine development against K. pneumoniae.

The group's latest software development introduces machine learning into the workflow. NextShift, developed by UCT Masters student Unays Bhad under the supervision of Kuttel and Dr Jan Buys, predicts nuclear magnetic resonance (NMR) chemical shifts from three-dimensional carbohydrate structures. The work builds on carbohydrate NMR prediction research involving Professor Göran Widmalm and collaborators at Stockholm University. Such predictions can assist with interpreting the complex experimental data used to determine carbohydrate structures. NextShift helps researchers assess proposed structures and compare alternative interpretations. Honours students Tanweer Chiktay and Fayyaaz Harnekar are currently integrating it into SugarDB, linking carbohydrate structures with predicted NMR data. Their work reflects the cumulative nature of the programme, in which successive student projects extend and connect earlier tools.

CarbBuilder, SugarDB, NextShift and CarbVis support connected stages of carbohydrate research. The long-term goal is to let researchers search for structures, generate three-dimensional models, predict NMR data and visualise molecular behaviour within an integrated workflow.

For Kuttel, the software itself represents an important research contribution. "Scientific knowledge is embedded in the algorithms, data structures and visual representations built into the software," she explained. By carrying expertise forward across multiple student projects and collaborations, the group is creating reusable research infrastructure that supports future discovery in glycoscience, vaccine development and beyond.

Lead researcher and contact: Professor Michelle Kuttel, Department of Computer Science, University of Cape Town — mkuttel@cs.uct.ac.za.