Title: Identification of Cancer Patient Subgroups via Smoothed Shortest Path Graph Kernel

Abstract: Characterizing patient somatic mutations through next-generation sequencing tech-nologies opens up possibilities for refining cancer subtypes. However, cataloguesof mutations reveal that only a small fraction of the genes are altered frequentlyin patients. On the other hand different genomic alterations may perturb the samepathways. We propose a novel clustering procedure that quantifies the similaritiesof patients from their mutational profile on pathways via a novel graph kernel.We represent each KEGG pathway as an undirected graph. For each patient thevertex labels are assigned based on her altered genes. Smoothed shortest pathgraph kernel (smSPK) evaluates each pair of patients by comparing their vertexlabeled pathway graphs. Our clustering procedure involves two steps: the smSPKkernel matrix derived for each pathway are input to kernel k-means algorithmand each pathway is evaluated individually. In the next step, only those pathwaysthat are successful are combined in to a single kernel input to kernel k-means tostratify patients. Evaluating the procedure on simulated data showed that smSPKclusters patients up to 88% accuracy. Finally to identify ovarian cancer patientsubgroups, we apply our methodology to the cancer genome atlas ovarian datathat involves 481 patients. The identified subgroups are evaluated through survivalanalysis. Grouping patients into four clusters results with patients groups that aresignificantly different in their survival times (p-value \leq 0.005).