Uncovering key transcription factors in breast cancer subtypes using matrix factorization

Breast cancer is the most common cancer type in women, and response to treatment varies immensely between subtypes. As of today, patients with Basal- like breast cancer lacks targeted treatment, which leads to poor prognosis for this group. Also other subtypes could benefit from a more targeted treatment. The molecular characteristics of each subtype remains an active area of research, and transcription factors that drive the subtypes need to be investigated in order to provide potential targets for more effective treatments. The molecular characteristics of each breast cancer subtype were inferred from ATAC-seq and RNA-seq data from 70 breast cancer patients, using two different matrix factorization methods. The first analysis used non-negative matrix factorization (NMF) on two separate data sets: One for ATAC-seq data, and one for RNA-seq data. The samples were clustered into five groups, based on molecular patterns shared within the groups, for both data sets. The DNA regions that were specifically open for each group were investigated for enriched transcription factor binding sites. The same was done for the promoter regions of the genes that were highly expressed in each group. The Basal-like subtype achieved the most successful clustering, and transcription factors likely to drive this subtype were uncovered. Also transcription factors responsible for driving a collective group of estrogen positive (ER+) subtypes were uncovered. The second analysis used Multi-Omics Factor Analysis (MOFA) to integrate the ATAC-seq and RNA-seq data in one combined analysis. The main purpose of this analysis was to support the findings of the first analysis, and possibly improve the clustering. The integration of multi- omics data resulted in two clusters, separating the Basal-like subtype from the rest of the subtypes. The clustering was not improved. However, some of the key transcription factors found for each group supported the results of the NMF analysis