Identifying miRNA synergism using multiple-intervention causal inference

Junpeng Zhang, Vu Pham, Lin Liu, Taosheng Xu, Buu Truong, Jiuyong Li, Nini Rao, Thuc Duy Le

TL;DR

This paper introduces miRsyn, a novel computational framework that uses multiple-intervention causal inference to identify synergistic interactions between microRNAs (miRNAs) from both sequence and expression data. This approach better reflects biological reality than previous methods, revealing that many miRNA pairs with shared targets at the sequence level do not act synergistically at the expression level, and highlighting the importance of these synergistic interactions in breast cancer.

What Problem Does This Paper Address?

Understanding how multiple microRNAs (miRNAs) work together, or synergize, is crucial for deciphering the complex regulatory mechanisms underlying human diseases, particularly cancers. Current methods often assume that miRNAs sharing target genes at the sequence level are inherently synergistic. However, this assumption is questionable, as these miRNAs might act independently or at different times. The gold standard for testing synergy involves simultaneously knocking down multiple miRNAs, which is experimentally challenging due to the sheer number of combinations. This paper addresses the need for a more practical and accurate computational method to identify miRNA synergism, especially considering the simultaneous action of multiple miRNAs.

Methods

The core of this work is the miRsyn framework, which employs a multiple-intervention causal inference approach. Instead of direct experiments, miRsyn simulates "multiple-intervention experiments" (like knocking down several miRNAs at once) using observational data. The process involves:

  1. Feature Selection: Identifying significant miRNAs and mRNAs from expression data using a Cox regression model.
  2. Causal Inference: Estimating "joint causal effects" between significant miRNAs and mRNAs using a parallelized Joint-IDA algorithm, which is an extension of causal inference methods designed for high-dimensional data. This is a key novelty, as it moves beyond single-intervention causal inference.
  3. Network Construction: Integrating causal effects with putative miRNA-target binding information (from miRTarBase) to build a miRNA synergistic network. This involves identifying sets of miRNAs with maximum joint causal effects on target mRNAs.
  4. Module Identification: Finding "bi-cliques" (sets of miRNAs and mRNAs) and then identifying synergistic modules within these bi-cliques based on maximum cumulative joint causal effects.
  5. Functional Analysis: Performing topological analysis (e.g., small-world properties) and functional enrichment analysis (e.g., Gene Ontology, pathways, diseases) to understand the biological significance of the identified synergistic networks and modules, with a specific focus on breast cancer.

Datasets

The study utilizes the following datasets:

  • The Cancer Genome Atlas (TCGA) breast cancer dataset: This provided matched miRNA and mRNA expression data, as well as clinical information for breast cancer samples.
  • miRTarBase v7.0: Used for experimentally validated miRNA-target interactions.
  • HMDD v3.0, miR2Disease, miRCancer, oncomiRDB: Databases for breast cancer-related miRNAs.
  • DisGeNET v5.0, COSMIC v86: Databases for breast cancer-related genes.
  • PmmR database: Used for validating predicted miRNA-miRNA synergistic pairs.

No new benchmarks were introduced; rather, existing databases and datasets were leveraged.

Key Insights and Ideas

  • Sequence vs. Expression Level Synergy: A critical finding is that many miRNA-miRNA pairs that appear to have synergistic interactions based on shared targets at the sequence level do not exhibit synergistic behavior at the expression level. This highlights the limitations of purely sequence-based predictions and the necessity of integrating expression data for accurate synergy identification.
  • Multiple Interventions Matter: The paper strongly advocates for and demonstrates the superiority of multiple-intervention causal inference over single-intervention methods. This is because biological synergy often involves the coordinated action of several miRNAs simultaneously.
  • Small-World Networks: The identified miRNA synergistic network exhibits "small-world" properties (dense local neighborhoods and short average path lengths). This characteristic is thought to facilitate rapid adaptation to environmental changes, which is relevant for biological systems responding to disturbances.
  • Biologically Meaningful Findings: The identified synergistic networks and modules are not just statistically significant but also biologically relevant, showing enrichment in processes and pathways associated with breast cancer.

Key Results

  • miRsyn identified 702 miRNA-miRNA synergistic pairs involving 78 miRNAs.
  • 361 miRNA synergistic modules were identified, with 72 significantly enriched in breast cancer.
  • ~71.08% of synergistic miRNA-miRNA pairs showed similar expression patterns, suggesting coordinated behavior.
  • Comparison with mirSRN (a single-intervention method): miRsyn predicted more synergistic pairs (702 vs. 239) and showed better performance in terms of functional enrichment, network topology (higher clustering coefficient, shorter characteristic path length), and overlap with high-confidence pairs from the PmmR database.
  • The study found that 611 out of 1313 miRNA-miRNA pairs with shared targets at the sequence level were not synergistic at the expression level, underscoring the importance of expression data.

Limitations and Future Work

  • Validation: While the study uses databases like PmmR for validation, the authors acknowledge that a definitive "ground-truth" for miRNA-miRNA synergistic pairs is still lacking.
  • Comparison Scope: The comparison was primarily focused on one existing method (mirSRN). The authors suggest that future work should involve comparing miRsyn to a broader range of methods to further demonstrate its performance.
  • Data Heterogeneity: The paper emphasizes the need to study miRNA synergism from heterogeneous data sources, implying that integrating even more diverse data types could yield further insights.
  • Generalizability: While the framework is presented as applicable to other complex diseases, further validation in different disease contexts would be beneficial.

Summary updated

Abstract

Abstract Background Studying multiple microRNAs (miRNAs) synergism in gene regulation could help to understand the regulatory mechanisms of complicated human diseases caused by miRNAs. Several existing methods have been presented to infer miRNA synergism. Most of the current methods assume that miRNAs with shared targets at the sequence level are working synergistically. However, it is unclear if miRNAs with shared targets are working in concert to regulate the targets or they individually regulate the targets at different time points or different biological processes. A standard method to test the synergistic activities is to knock-down multiple miRNAs at the same time and measure the changes in the target genes. However, this approach may not be practical as we would have too many sets of miRNAs to test. Results In this paper, we present a novel framework called miRsyn for inferring miRNA synergism by using a causal inference method that mimics the multiple-intervention experiments, e.g. knocking-down multiple miRNAs, with observational data. Our results show that several miRNA-miRNA pairs that have shared targets at the sequence level are not working synergistically at the expression level. Moreover, the identified miRNA synergistic network is small-world and biologically meaningful, and a number of miRNA synergistic modules are significantly enriched in breast cancer. Our further analyses also reveal that most of synergistic miRNA-miRNA pairs show the same expression patterns. The comparison results indicate that the proposed multiple-intervention causal inference method performs better than the single-intervention causal inference method in identifying miRNA synergistic network. Conclusions Taken together, the results imply that miRsyn is a promising framework for identifying miRNA synergism, and it could enhance the understanding of miRNA synergism in breast cancer.
BibTeX
@article{zhang2019identifying,
  author = {Junpeng Zhang and Vu Pham and Lin Liu and Taosheng Xu and Buu Truong and Jiuyong Li and Nini Rao and Thuc Duy Le},
  title = {{Identifying miRNA synergism using multiple-intervention causal inference}},
  journal = {BMC Bioinformatics},
  year = {2019},
  doi = {10.1186/s12859-019-3215-5},
}
Figure