Modeling miRNA-mRNA interactions that cause phenotypic abnormality in breast cancer patients

Abstract

Background: The dysregulation of microRNAs (miRNAs) alters expression level of pro-oncogenic or tumor suppressive mRNAs in breast cancer, and in the long run, causes multiple biological abnormalities. Identification of such interactions of miRNA-mRNA requires integrative analysis of miRNA-mRNA expression profile data. However, current approaches have limitations to consider the regulatory relationship between miRNAs and mRNAs and to implicate the relationship with phenotypic abnormality and cancer pathogenesis.

Methodology/findings: We modeled causal relationships between genomic expression and clinical data using a Bayesian Network (BN), with the goal of discovering miRNA-mRNA interactions that are associated with cancer pathogenesis. The Multiple Beam Search (MBS) algorithm learned interactions from data and discovered that hsa-miR-21, hsa-miR-10b, hsa-miR-448, and hsa-miR-96 interact with oncogenes, such as, CCND2, ESR1, MET, NOTCH1, TGFBR2 and TGFB1 that promote tumor metastasis, invasion, and cell proliferation. We also calculated Bayesian network posterior probability (BNPP) for the models discovered by the MBS algorithm to validate true models with high likelihood.

Conclusion/significance: The MBS algorithm successfully learned miRNA and mRNA expression profile data using a BN, and identified miRNA-mRNA interactions that probabilistically affect breast cancer pathogenesis. The MBS algorithm is a potentially useful tool for identifying interacting gene pairs implicated by the deregulation of expression.

Fig 1. A BN representing relationships among variables related to prostate cancer.

Fig 2. The model for the effect of miRNA-i on the target through the hidden variable.

(a). The model including the hidden variable H, (b) The interaction model learned from data.

Fig 3. Schematic diagram for the pre-processing workflow and learning DAG models on microRNA and mRNA expression profile data.

Fig 4. The expression level of hsa-miR-96 and its target mRNAs that are associated with breast cancer.

The expression levels of miRNAs and mRNAs were represented in 76 normal tissue samples and 747 tumor patients. t-test p-values are indicated above the box plots.