Medulloblastoma (MB), the most common malignant pediatric brain cancer, has four subtypes based on genomics profiling: Wingless (WNT), Sonic Hedgehog (SHH), Group 3, and Group 4. However, subtype heterogeneity and low mutation rates hinder the identification of actionable molecular targets within each group. Studies thus suggest a phosphoproteomics (study of protein phosphorylation), instead of a genomics approach, to better distinguish the MB subgroups. Thus, this research will measure MB’s irregular protein activity via protein phosphorylation patterns. Protein kinases phosphorylate proteins on specific amino acids, also known as substrates, to regulate their functions. Given that less than 5% of substrates have annotated kinase associations, it is difficult to identify the kinases in MB which irregularly phosphorylate their substrates in order to determine the most effective kinases for drugs to target. As a result, the publicly-available tool NetworKIN was employed to predict novel kinase-substrate interactions in this study to expand the currently limited annotated interactions. For a given dataset of substrates, NetworKIN will output kinasesubstrate predictions, each receiving a confidence score (higher score $=$ higher confidence) ranging from zero to the hundreds. To identify a score threshold for high confidence predictions, NetworKIN was run on the substrates in the Human PhosphoSite Plus Database (PSP), the largest curated database of experimentally verified kinase-substrate interactions in humans. Three statistical thresholding methods were tested and the optimal cutoff score was determined to lie at 5.88. This threshold was applied to the predictions for the MB dataset, resulting in 774 novel high confidence kinase-substrate interactions. To understand how these interactions altered the enriched kinases across the MB subgroups, kinase activity scores before and after adding the predictions were calculated for all 20 samples using the tool Inference of Kinase Activities from Phosphoproteomics. Subgroup classification based on these activity scores revealed that, with the addition of high-confidence interactions to PSP, there are two smaller subgroups of Group 3 and one of those subgroups more closely resembled the Group 4 samples. The alignment of these findings with current literature increases our confidence in the prediction algorithm and elucidates novel phosphorylation patterns to more accurately portray the phosphosignaling pathways driving MB tumorigenesis. Future cancer phosphoproteomics studies can also employ this robust prediction tool to continue expanding our understanding of aberrant kinase-substrate interactions in human diseases.