Alisha Patel
Pronouns: she/her
Research Mentor(s): Cristian Minoccheri
Research Mentor School/College/Department: Computational Medicine and Bioinformatics / Medicine
Program: UROP
Session: Session 5 (2:40pm – 3:30pm)
Authors: Alisha Patel, Cristian Minoccheri
Abstract
Multi-omics data is often analyzed to enrich our understanding of biological traits based on underlying phenotypes. However, the high-order correlations and the high dimension of this data make it difficult to affiliate certain omics features to a particular biological mechanism. For example, cancer subtypes may be well understood at the molecular level, but a challenge arises when classification must occur using gene expression data. In this project, we apply tensor methods to multi-omics analysis as applied to cancer subtyping. The multi-omics data are integrated into a three-dimensional tensor, which is comprised of gene expression, methylation, and miRNA. We will use algorithms to perform integrative analysis of clusters in multi-omic data to reveal if there are correlations between genetic markers and the expression of cancer subtypes. This study is important as analyzing these clusters will be able to give way to how the presence of a particular gene may have an impact on a person’s chance to develop cancer. This kind of analysis has been done before, but mostly at the matrix level. Even when using tensor methods, data has to be preprocessed to obtain matrices of the same size with two common modes for genes and patients (multi-staged integration). Parafac2 allows the integration of the data as is (multi-dimensional integration) with only one common mode for patients. Parafac2 has been used to extract phenotypes from electronic hospital records (EHR) data, but not omic data. The goal is to see whether Parafac2 can lead to novel results and/or outperform other methods.
