Computational integrative genomics approach to refine breast cancer genetic grading and prognosis
Date of Issue2016-09-26
School of Computer Engineering
Bioinformatics Research Centre
Agency for Science, Technology and Research (A*STAR) , Bioinformatics institute (BII)
Tumor heterogeneity is one of the major challenges of breast cancer (BC) classification and patients’ treatment. Histologic grading classifies BC into three categories: 1, 2 and 3 (HG1, HG2, and HG3, respectively). HG has so far been one of the most powerful BC prognostic markers for patients’ survival. However, it has several limitations, such as the moderate inter-observer reproducibility and poor prognostic value for HG2, which consists of 50% of BC patients. Previous transcriptome data analyses allowed the identification of gene signatures that best discriminate between HG1 and HG3 tumors. These genetic signatures were also able to re-classify HG2 tumors into two sub-classes, termed HG1-like and HG3-like tumors, constitute more than 95% of HG2 tumor samples. HG1-like and HG3-like tumors were similar to HG1 and HG3 tumors respectively at the molecular and prognostic level. To improve BC classification and patients’ grading, I refined the grading concept of BC by performing integrative genomics analysis of BC histological and genetic grades concerning cancer cell origin, mutations, chromosome, and gene expression alterations and aimed to identify the oncogenic pathways governing BC classes development and progression. Using The Cancer Genome Atlas (TCGA) cohort data and other datasets, I conducted integrative genomics data analysis to study the low- and high- grade molecular classification of invasive ductal carcinoma (IDC), the major histologic type of BC. I tested the hypothesis that the low- and high- molecular grade classes are pre-determined by different cellular phenotypes and genetic programs driven by specific subsets of somatic mutations, DNA alterations and gene expression patterns delineating distinct oncogenic pathways. Using transcriptome data analysis, I identified a 22-genes signature that reclassifies HG2 tumors. I demonstrated that this classification was concordant with specific patterns of chromosome breaks and unique subsets of point mutations. Collectively, this analysis demonstrated the lack of genetic basis of IDC HG2 and further reclassified IDC tumors into two major genetic grades: low genetic grade (LGG=HG1+HG1-like) and high genetic grade (HGG= HG3-like+HG3) tumors. This study demonstrated that basal, ERBB2, and luminal-B tumor subtypes progress along the HGG oncogenic pathway, while normal-like and luminal-A tumors progress along the LGG oncogenic pathway. LGG and HGG were also distinct in stem cell-related genes expression profiles and predicted relatively favourable and unfavourable disease outcome respectively. Within LGG and HGG tumors, this computational integrative genomics study revealed relatively low and high-risk of disease subclasses pre-selected by specific genome instability regions that were associated with different patients’ prognosis and therapeutic responsiveness. I identified intra-genetic grades signatures that reflect different tumor proliferation and tumors-milieu interaction characteristics, including tumor immunity. This study suggests a new BC classification model that is able to stratify IDC into four (or more) prognostic groups correlating significantly with specific, molecular profiles, and risk of disease recurrence and prediction. This study could help in proposing a new avenue for future studies of breast carcinoma pathogenesis, optimization of systemic therapy and the discovery of clinically relevant biomarkers.
DRNTU::Engineering::Computer science and engineering::Computing methodologies::Pattern recognition