KSFinder – a Model to Predict Kinases of Phosphorylated Substrates, Its Enhancement, and Applications in Human Diseases
Manju Anandakrishnan
PhD Student, Bioinformatics Data Science Program
CBCB, University of Delaware
Abstract: Aberrant protein kinase regulation leading to abnormal substrate phosphorylation is associated with several human diseases. Despite the promise of drug therapies targeting kinases, for a large proportion of human phosphosites, catalyzing kinase is unknown. Kinase inference methods are techniques that use differential phosphorylation abundance of substrates to determine implicated kinases in conditions. They work with the limited known kinase-substrate association data or expand their background with supplemental protein interaction data. Our goal is to augment the kinase-substrate association data using computational predictions. Most existing computational tools predicting kinase-substrate relation cover less than 50% of known human kinases, and several utilize manual feature selection based on protein sequences, motifs, domains, structures, and/or functions. Using heterogeneous relationships of proteins in a cellular network, and a combinatorial negative generation strategy, we built KSFinder, a knowledge graph embedding and deep learning-based kinase-substrate (protein-level) prediction model that exhibited generalization ability on different datasets and outperformed other SOTA models. Expanding on this approach, we are developing KSFinder 2.0 to predict kinase-substrate associations at the site level, using sophisticated feature extraction techniques – knowledge graph embedding and transformer-based models, which aim to extract features based on protein sequence, structure, and functional relationships.
Bio: Manju is a Ph.D. student in the Bioinformatics Data Science program. She is doing her research in predicting kinase-substrate phosphorylation using advanced machine-learning techniques under the supervision of Dr. Cathy Wu. Manju received her bachelor’s degree in Genetic Engineering from SRM University, India. Before joining UD, she worked in the field of software engineering for 12 years. Her research interests include applying machine learning techniques and developing software tools to study molecular mechanisms driving human diseases.
