|
Abstract:
|
Cancer subtypes discovery is the first step to deliver personalized medicine to cancer patients. With the accumulation of massive multi-omics datasets and established biological knowledge databases, omcis data integration with incorporation of rich existing biological knowledge is essential for deciphering biological mechanism behind the complex diseases. In this talk, we propose an integrative sparse K-means (IS-Kmeans) approach to combine multi-omics datasets to discover disease subtypes with the guidance of prior biological knowledge via a sparse overlapping group lasso technique. A fast algorithm of alternating direction method of multiplier (ADMM) will be applied to optimized the proposed objective function. Simulation and real data application in breast cancer will be used to compare IS-Kmeans with other existing methods and demonstrate its superior clustering accuracy, computing efficiency and functional annotation of detected molecular features.
|
