Prof. Dehne is a distinguished researcher in data science and parallel computing. Most of his work is interdisciplinary and collaborative. He specializes in parallel algorithms design for cloud computing, multi-core processors, and GPUs to create high-performance computing systems for solving data-intensive and computationally hard problems in bioinformatics, business data analytics, and computational engineering.
Bioinformatics
Together with Professor Ashkan Golshani from the Institute for Biochemistry, Prof. Dehne leads an interdisciplinary research group on computational proteomics and computational drug design. The group developed the first high-performance computing system (named PIPE) for high-precision protein to protein interaction prediction and the design of new synthetic peptides with specific interaction properties. Their computational method is based on a novel machine learning algorithm that utilizes in-depth knowledge from experimental protein to protein interaction detection results in Professor Golshani’s lab. PIPE has won international acclaim and has enabled new biomedical research. They collaborated with the Ottawa Hospital Research Institute on new synthetic peptides for muscular dystrophy stem cell therapy; with the Department of Biochemistry at the University of Regina on designing anti Zika virus peptides; with the School of Medicine at WITS University in Johannesburg on new methods to explain evolutionary transitions; and with Agriculture Canada on the development of new soy bean plants that withstand cold climates. In collaboration with the Ottawa Hospital Research Institute and the Department of Biochemistry at the University of Regina, they designed, via large-scale computation, a new synthetic peptide to block the interaction between the COVID spike protein and the human ACE2 receptor. Their novel peptide was synthesized and tested at the Ottawa Hospital Research Institute and the National Microbiology Laboratory in Winnipeg (on live COVID viruses). Both lab tests were successful.
Business Data Analytics
Prof. Dehne is interested in interdisciplinary research projects that involve large data sets and computationally hard problems. One such example is large scale business data analytics (“big data”). IBM purchased Cognos Corporation in Ottawa and established Ottawa as IBM’s main centre for data analytics in Canada. When installing big data systems for their clients, IBM encountered technical difficulties because their software tools had performance issues on very large data sets. That lead to discussions between IBM/Cognos and Prof. Dehne’s research lab at Carleton University. The data analytics operations that created these performance issues were complex aggregate (group-by) queries on large data sets, with the additional challenge that those data sets were highly dynamic. IBM funded a research project in Prof. Dehne’s lab to address this problem. Their solution and delivered prototype won an IBM Innovation Impact Of The Year Award, and Prof. Dehne was appointed Fellow of the IBM Centre For Advanced Studies in Toronto, Canada.
Computational Welding Mechanics
Computational welding mechanics is about the precise simulation of welding processes with the goal of controlling welding robots to perform complex welding tasks. The field was pioneered by Professor John Goldak who is also the principal author of the eminent book. The computations are extremely complex and time consuming. Prof. Dehne collaborated with Prof. Goldak on how to parallelize his welding simulation algorithms and execute them on arrays of GPUs (e.g. five GPUs with 4,000 processor cores each). That enabled Prof. Goldak to greatly improve the speed and precision of his welding simulations, leading to the commercial development of new real-time, high precision, multi-GPU control units for welding robots. These control units are now routinely used by Canadian oil companies for robotic welding of oil pipelines.