WONG Limsoon first made his name in the database theory community for a series of fundamental results on the expressive power of nested relational languages augmented with aggregate functions. His theorems on the conservative extension property and the bounded degree property of these languages were breakthroughs that settled some long-standing issues in understanding the expressive power of practical query languages such as SQL. One paper from this series of work was recognized by the ICDT 2014 Test of Time Award.  Project webpage

His second well-known contribution to database theory was his role in the solution of the Kanellakis Conjecture on the inexpressibility of transitive closure in a large class of constraint query languages. He made an insightful and fundamental conjecture that this class of query languages under the natural semantics interpretation would collapse to the active semantics interpretation for all queries satisfying a simple genericity condition and proceeded to confirm this conjecture in a collaboration with Michael Benedikt, Leonid Libkin, and Guozhu Dong in a much-celebrated paper. The long-standing Kanellakis Conjecture followed as a direct corollary. Project webpage

Limsoon’s third important contribution was the development of the Kleisli Query System for broad-scale data integration. The system was the first to solve a series of data integration problems considered “impossible” by the US Department of Energy at that time. This seminar work was also considered to be one of the six top practical achievements of the 1990’s in functional programming by that research community as announced in Wadler’s 1998 “Angry Half Dozen” article in SIGPLAN Notices. His work on Kleisli also won him one of the only two Tan Kah Kee Young Inventor Gold Awards (in the open category) ever given in the circa 40 years’ history of the competition series. Project webpage

His next significant contribution was his leadership in establishing natural language processing (NLP) as an important topic in bioinformatics. He initiated the first dedicated bioinformatics conference track on this topic and organized the first three meetings. He wrote one of the first two papers on building an NLP system for extracting protein interaction and regulation information from literature. He also lead-authored the key paper that called the community into action to establish a series of benchmarking competitions on this topic. He co-founded (with See-Kiong Ng) the first commercial company (Molecular Connections Pvt Ltd) that offered services on this topic. The profitable company has grown 400x in two decades. It now has about 2,700 engineers, scientists, and curators and is one of the best SME’s in India. Project webpage

From the early 2000s, Limsoon switched his attention from database theory, data integration, and information extraction to problems in computational biology that require profound domain knowledge in biology and biotechnology. He is among the group of visionary computational biologists to actively pioneer knowledge discovery approaches that deeply incorporate biological background—e.g. biological pathways and networks, often integrated from multiple sources and extracted from free text—to model and solve problems in computational biology. A study reported in CIKM2006 ranked Limsoon as a top-10 social actor who powered data mining research into computational biology. He has three well-known series of works in this area. The first was the extremely widely cited work of Limsoon and his team (Liu Huiqing, Li Jinyan, and Allen Yeoh) with St Jude’s that pioneered knowledge discovery from patient gene expression profiles of childhood acute lymphoblastic leukemias (ALL).  It was the first gene expression profile-based work to distinguish all major pediatric ALL subtypes and the first to demonstrate the use of gene expression profiling to identify treatment relapse in certain childhood ALL subtypes. They also presented a novel subtype which is now recognized as the DUX4 subtype—the largest B-lineage ALL subtype among patients with poor initial treatment responses, yet yielding excellent eventual survival outcomes. This work won the team the FEER Asian Innovation Gold Award in 2003.  Project webpage

The second was his seminar work (with Hon Nian Chua) on the cleansing and analysis of protein interaction networks. They made a discovery that proteins have much stronger functional association with their indirect neighbours than their immediate neighbours in protein-protein interaction networks. This was totally unexpected by both wet and computational biologists. Limsoon’s team successfully and elegantly exploited this phenomenon to make reliable protein function prediction even in the absence of sequence homology.  His team’s key paper on this topic (“Exploiting indirect neighbours and topological weight to predict protein function from protein-protein interactions”, Bioinformatics, 2006) was rated by ISI as the 49th hottest computer science paper published in 2006. Also, his team was winner of the DREAM2 Challenge in 2007 on this topic. Project webpage

The third and most recent is an ongoing series of work (with Wilson Goh and Weijia Kong) on analyzing proteomics datasets which have severe reproducibility and coverage issues. They insightfully pointed out that, for disease phenotypes caused by the non-functioning of a protein complex, a major change in any protein component of the complex would lead to the non-functioning of the complex and, thus, the disease phenotype. Therefore, while one often could not see a consistent proteomic profile at the level of individual proteins (since different ones could change in different samples), one would see a consistent proteomic profile at the level of protein complexes.  This insight has led to successful pioneering protein complex-based approaches to overcome, in simple yet refined ways, the consistency and coverage challenges that have plagued untargeted proteomic profiling. Project webpage