Clustering Tree-Structured Data on Manifold

Elastic Differential Evolution for Automatic Data Clustering

In many practical applications, it is crucial to perform automatic data clustering without knowing the number of clusters in advance. The evolutionary computation paradigm is good at dealing with this task, but the existing algorithms encounter several deficiencies, such as the encoding redundancy and the cross-dimension learning error. In this article, we propose a novel elastic differential evolution algorithm to solve automatic data clustering. Unlike traditional methods, the proposed algorithm considers each clustering layout as a whole and adapts the cluster number and cluster centroids inherently through the variable-length encoding and the evolution operators. The encoding scheme contains no redundancy. To enable the individuals of different lengths to exchange information properly, we develop a subspace crossover and a two-phase mutation operator. The operators employ the basic method of differential evolution and, in addition, they consider the spatial information of cluster layouts to generate offspring solutions. Particularly, each dimension of the parameter vector interacts with its correlated dimensions, which not only adapts the cluster number but also avoids the cross-dimension learning error. The experimental results show that our algorithm outperforms the state-of-the-art algorithms that it is able to identify the correct number of clusters and obtain a good cluster validation value.

Integration of Cancer Genomics Data for Tree-based Dimensionality Reduction and Cancer Outcome Prediction

Accurate outcome prediction is crucial for precision medicine and personalized treatment of cancer. Researchers have found that multi-dimensional cancer omics studies outperform each data type (mRNA, microRNA, methylation or somatic copy number alteration) study in human disease research. Existing methods leveraging multiple level of molecular data often suffer from various limitations, e. g., heterogeneity, poor robustness or loss of generality. To overcome these limitations, we presented the tree-based dimensionality reduction approach for the identification of smooth tree graph and developed accurate predictive model for clinical outcome prediction. We demonstrated that 1) Discriminative Dimensionality Reduction via learning a Tree (DDRTree) achieved reduced dimension space tree with statistical significance; 2) Tree based support vector machine (SVM) classifier improved prediction performance of cancer recurrence as compared to t-test based SVM classifier; 3) Tree based SVM classifier was robust with regard to the different number of multi-markers; 4) Combining multiple omics data improved prediction performance of cancer recurrence as compared to a single-omics data; and 5) Tree based SVM classifier achieved similar or better prediction performance when compared to the features from state-of-the-art feature selection methods. Our results demonstrated great potential of the tree-based dimensionality reduction approach based clinical outcome prediction.