Abstract 2671
Background
Homeobox (HOX) family consists of 39 genes which act as master regulators in embryonic development. Each of the genes is also known to play key roles in progression of breast cancer, including epithelial to mesenchymal transition, tumor angiogenesis and endocrine therapy resistance. Although there are numerous reports on individual HOX genes and cancer, none of them have comprehensively analyzed the whole gene family. Since HOX genes strongly coordinate within the family during the embryonic period, we considered that the analysis of the whole HOX family is also indispensable in breast cancer.
Methods
We collected 702 breast cancer data from four publicly available array datasets (GSE11121, GSE7390, GSE3494, GSE2990) and performed unsupervised hierarchal clustering into two clusters by the expression of HOX genes. We constructed model formulas for cluster prediction by dividing the samples into learning and validation groups. We used three machine learning methods: support-vector machine (SVM), neural network and Bayes. The model formulas were validated by validation samples. We also used 512 TCGA breast cancer data to calculate covariations of the genes in breast cancer.
Results
By the clustering of four array datasets, the DFS of the two clusters in PAM50-classified luminal B patients were statistically different (p = 0.016), and the gene ontology analysis revealed that the Wnt pathway was activated in the poor prognostic cluster. All cluster prediction models for luminal B sample achieved accuracies of over 90%. From TCGA breast cancer data, we found that HOX genes covariate the most with other HOX genes, especially within the chromosomally proximal groups.
Conclusions
Comprehensive analysis of the whole HOX family lead to the prediction of luminal B breast cancer prognosis. Considering that Wnt signaling controls HOX genes during the embryonic stage, we suppose a Wnt pathway activated, poor prognostic subgroup in luminal B breast cancer which can be identified by the expression of HOX genes. The cluster prediction model by machine learning was acceptable for its future adaptation in clinical settings. We also proved that HOX genes strongly covariate within the gene family in cancer, not only during the embryonic stage.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.
Resources from the same session
3664 - Longitudinal changes in cell-free DNA (cfDNA) methylation levels identify early non-responders to treatment in advanced solid tumors
Presenter: Andrew Davis
Session: Poster Display session 3
Resources:
Abstract
3212 - Multigene panel testing results for hereditary breast cancer in 1325 individuals: implications for gene selection and considerations for guidelines.
Presenter: Georgios Tsaousis
Session: Poster Display session 3
Resources:
Abstract
2591 - PIK3R5 genetic predictors of hypertension induced by VEGF-pathway inhibitors
Presenter: Julia Quintanilha
Session: Poster Display session 3
Resources:
Abstract
4377 - ERBB2 mRNA as a predictor in HER2-positive (HER2+)/hormone receptor-positive (HR+) metastatic breast cancer (BC) treated with HER2 blockade in combination with endocrine therapy (ET): a retrospective analysis of the ALTERNATIVE and SOLTI-PAMELA trials.
Presenter: Nuria Chic
Session: Poster Display session 3
Resources:
Abstract
3439 - Early on-treatment vs pre-treatment tumor transcriptomes as predictors of response to neoadjuvant therapy for HER2-positive inflammatory breast cancer
Presenter: Sonia Pernas
Session: Poster Display session 3
Resources:
Abstract
2512 - AXL expression predicts poor prognosis and lack of efficacy of anti-angiogenic and anti-epidermal growth factor receptor (EGFR) agents in patients (pts) with RAS wild type (WT) metastatic colorectal cancer (mCRC)
Presenter: Claudia Cardone
Session: Poster Display session 3
Resources:
Abstract
4061 - Prevalence of EGFR mutations and its correlation with Egyptian patients’ human kinetics (PEEK Study)
Presenter: Adel Ibrahim
Session: Poster Display session 3
Resources:
Abstract
2547 - Evaluation of tumor microenvironment identifies immune correlates of response to combination immunotherapy with margetuximab (M) and pembrolizumab (P) in HER2+ gastroesophageal adenocarcinoma (GEA)
Presenter: Sergio Rutella
Session: Poster Display session 3
Resources:
Abstract
4671 - Clinicopathological and molecular criteria assessment for the screening of hypermutated proficient mismatch repair (pMMR) colorectal cancers (CRC) with exonucleasic domain POLE (edPOLE) mutations (mt).
Presenter: Benoit Rousseau
Session: Poster Display session 3
Resources:
Abstract
3862 - Tumor mutation burden and microsatellite instability in colorectal cancer
Presenter: Francesca Fenizia
Session: Poster Display session 3
Resources:
Abstract