Identifying miRNA-mRNA regulatory relationships in breast cancer with invariant causal prediction

Vu Pham, Junpeng Zhang, Lin Liu, Buu Truong, Taosheng Xu, Trung T. Nguyen, Jiuyong Li, Thuc Duy LebioRxiv 2018 · DOI

TL;DR

This paper introduces a novel method for identifying miRNA-mRNA regulatory relationships in breast cancer by leveraging "invariant causal prediction" across different cancer subtypes. Their approach, particularly when combined with other methods in an ensemble, significantly outperforms existing techniques in predicting these crucial interactions, paving the way for better diagnostics and therapeutics.

What Problem Does This Paper Address?

MicroRNAs (miRNAs) are small RNA molecules that play a critical role in regulating gene expression, and their dysregulation is implicated in various diseases, including cancer. Identifying how miRNAs interact with messenger RNAs (mRNAs) to control gene expression is crucial for understanding cancer development and designing effective treatments. However, existing methods for predicting these miRNA-mRNA relationships have limitations: some only identify correlations, not true causal links, while others that do infer causality are computationally expensive and not scalable. This paper aims to bridge this gap by developing a more efficient and accurate method for discovering these regulatory relationships, specifically within the context of breast cancer subtypes.

Methods

The core of the paper's methodology is the application of Invariant Causal Prediction (ICP), a recently developed causal inference technique. The key idea behind ICP is that true causal relationships between a "cause" (miRNA) and an "effect" (mRNA) should remain consistent across different "environments." In this study, these "environments" are defined by different breast cancer subtypes, identified using the Pam50 classification method.

The authors adapt ICP, specifically a variant called hidden ICP, to account for potential unobserved factors (hidden miRNAs) that might influence the observed relationships. The process involves:

  1. Feature Selection: Identifying significant miRNAs and mRNAs based on their median absolute deviation in the breast cancer dataset.
  2. Environment Categorization: Using Pam50 to group samples into distinct cancer subtypes, serving as the different environments for ICP.
  3. Causal Relationship Estimation: Applying hidden ICP to find miRNA-mRNA pairs that exhibit consistent causal relationships across these identified subtypes.

To further enhance predictive power, they also developed an ensemble method that combines the proposed hidden ICP approach with traditional methods like Pearson correlation and Lasso regression. This ensemble leverages the strengths of each individual method.

Datasets

The primary dataset used in this study is the breast adenocarcinoma (BRCA) dataset from The Cancer Genome Atlas (TCGA), which contains matched miRNA and mRNA expression profiles.

For validation, the authors utilized:

  • miRNA-transfected experimental data (provided in S4 File).
  • Experimentally confirmed databases: miRTarbase version 6.1, TarBase version 7.0, and miRWalk version 2.0.

Key Insights and Ideas

  • Causality over Correlation: The paper strongly emphasizes the importance of identifying causal regulatory relationships rather than mere correlations, as correlations can be spurious and misleading in biological contexts.
  • Subtype-Specific Invariance: The concept of "invariant causal prediction" is ingeniously applied by using different breast cancer subtypes as distinct "environments." This allows the identification of regulatory relationships that are robust and consistent across the heterogeneity of breast cancer.
  • Ensemble Power: Combining a novel causal method with established statistical techniques (Pearson, Lasso) in an ensemble approach yields superior performance, highlighting the benefit of integrating diverse analytical perspectives.
  • Synergistic miRNA Action: The findings suggest that miRNAs often work together synergistically to regulate target genes, indicating a complex regulatory network that goes beyond individual miRNA-mRNA interactions.

Key Results

  • The proposed hiddenICP-Pam50 method significantly outperforms existing state-of-the-art methods in discovering miRNA-mRNA regulatory relationships when validated against transfection data.
  • The ensemble method, combining hidden ICP with Pearson and Lasso, demonstrated the best performance among all tested methods, even surpassing the individual causal method.
  • Functional enrichment analyses revealed that miRNAs involved in the predicted regulatory relationships tend to synergistically regulate target genes, suggesting their coordinated roles in biological processes.
  • The identified miRNA targets were found to be significantly enriched in functions and diseases related to breast cancer, underscoring the biological relevance of the findings.

Limitations and Future Work

  • Completeness of Transfection Data: The authors acknowledge that the miRNA transfection data used for validation is not exhaustive, which might limit direct comparisons for all possible miRNA-mRNA interactions.
  • Computational Complexity of Other Methods: While their proposed method is efficient, they note that other causal inference methods can have high computational complexity, making them less scalable for large datasets.
  • Future Work: The identified miRNA targets are proposed as candidates for future wet-lab experiments to further validate their roles and discover the underlying causes of breast cancer. The ensemble approach also suggests that further exploration of integrating different types of data and methods could yield even more robust predictions.

Summary updated

Abstract

Abstract microRNAs (miRNAs) regulate gene expression at the post-transcriptional level and they play an important role in various biological processes in the human body. Therefore, identifying their regulation mechanisms is essential for the diagnostics and therapeutics for a wide range of diseases. There have been a large number of researches which use gene expression profiles to resolve this problem. However, the current methods have their own limitations. Some of them only identify the correlation of miRNA and mRNA expression levels instead of the causal or regulatory relationships while others infer the causality but with a high computational complexity. To overcome these issues, in this study, we propose a method to identify miRNA-mRNA regulatory relationships in breast cancer using the invariant causal prediction. The key idea of invariant causal prediction is that the cause miRNAs of their target mRNAs are the ones which have persistent causal relationships with the target mRNAs across different environments. In this research, we aim to find miRNA targets which are consistent across different breast cancer subtypes. Thus, first of all, we apply the Pam50 method to categorise BRCA samples into different ‘‘environment” groups based on different cancer subtypes. Then we use the invariant causal prediction method to find miRNA-mRNA regulatory relationships across subtypes. We validate the results with the miRNA-transfected experimental data and the results show that our method outperforms the state-of-the-art methods. In addition, we also integrate this new method with the Pearson correlation analysis method and Lasso in an ensemble method to take the advantages of these methods. We then validate the results of the ensemble method with the experimentally confirmed data and the ensemble method shows the best performance, even comparing to the proposed causal method. Functional enrichment analyses show that miRNAs in the regulatory relationship predicated by the proposed causal method tend to synergistically regulate target genes, indicating the usefulness of these methods, and the identified miRNA targets could be used in the design of wet-lab experiments to discover the causes of breast cancer. Author summary Cancer is a disease of cells in human body and it causes a high rate of deaths world wide. There has been evidence that non-coding RNAs are key players in the development and progression of cancer. Among the different types of non-coding RNAs, miRNAs, which are short non-coding RNAs, regulate gene expression and play an important role in different biological processes as well as various cancer types. To design better diagnostic and therapeutic plans for cancer patients, we need to know the roles of miRNAs in cancer initialisation and development, and their regulation mechanisms in the human body. In this study, we propose algorithms to identify miRNA-mRNA regulatory relationships in breast cancer. Comparing our methods with existing methods in predicting miRNA targets, our methods show a better performance. The estimated miRNA targets from our methods could be a potential source for further wet-lab experiments to discover the causes of breast cancer.
BibTeX
@misc{pham2018identifying,
  author = {Vu Pham and Junpeng Zhang and Lin Liu and Buu Truong and Taosheng Xu and Trung T. Nguyen and Jiuyong Li and Thuc Duy Le},
  title = {{Identifying miRNA-mRNA regulatory relationships in breast cancer with invariant causal prediction}},
  year = {2018},
  doi = {10.1101/340638},
  howpublished = {bioRxiv},
}
Figure