Computer science research team gains international recognition for method that accurately predicts intrinsic disorder in proteins
Editor’s note: Kurgan’s award-winning method now appears in the journal Nature Communications. The article, titled “flDPnn: Accurate intrinsic disorder prediction with putative propensities of disorder functions,” was published July 21, 2021. The editors of Nature Communications also placed Kurgan’s article on the Editor’s Highlights page, which features a small selection of articles the editorial team believes to be particularly interesting or important.
A computer science research team from VCU Engineering won an international challenge for their novel method of predicting intrinsically disordered proteins. These proteins are inherently unstructured and have been found to be associated with cancers, cardiovascular and neurodegenerative diseases, which makes them promising targets for drug discovery.
Lukasz Kurgan, Ph.D., the Robert J. Mattauch Professor and vice chair of VCU’s Department of Computer Science, and a team of his doctoral students and collaborators won first place in Critical Assessment of Protein Intrinsic Disorder Prediction (CAID). This worldwide challenge was established as a community-based blind test to identify the most accurate methods that predict unstructured protein regions.
CAID complements the Critical Assessment of Protein Structure Prediction (CASP) challenge that evaluates methods that predict protein structure. Google’s AlphaFold recently won CASP.
The Kurgan team’s entry, called flDPnn, outperformed a record-breaking pool of 32 methods developed by research teams from across the world.
The results were adjudicated by a panel of international experts and were published in the journal Nature Methods, followed by a commentary article in the same journal that highlights the key findings.
VCU Engineering computer science doctoral students Akila Katuwawala and Sina Ghadermarzi, along with several of Kurgan’s collaborators from Nankai University in Tianjin, China, contributed to the development of this method. It predicts intrinsically disordered proteins using a deep neural network that relies on a sophisticated approach to encode network inputs derived from protein sequences. This method is a culmination of more than a decade of research in this area, most recently sponsored by the National Science Foundation.
“Developing successful bioinformatics tools is an inherently multidisciplinary effort. It requires in-depth understanding of the underlying molecular biology, knowledge of modern data science algorithms and principles and [the] engineering [of] complex software that makes the predictions,” Kurgan said. “I am blessed to work with a great team of talented people who came together to develop this novel tool. We are humbled by this recognition and highly motivated to repeat this achievement in the future editions of the CAID challenge.”
Categories Computer Science