OmicsARules: a R package for integration of multi-omics datasets via association rules mining

Danze Chen,Fan Zhang,Qianqian Zhao,Jianzhen Xu

doi:10.1186/s12859-019-3171-0

Abstract

BackgroundThe improvements of high throughput technologies have produced large amounts of multi-omics experiments datasets. Initial analysis of these data has revealed many concurrent gene alterations within single dataset or/and among multiple omics datasets. Although powerful bioinformatics pipelines have been developed to store, manipulate and analyze these data, few explicitly find and assess the recurrent co-occurring aberrations across multiple regulation levels.ResultsHere, we introduced a novel R-package (called OmicsARules) to identify the concerted changes among genes under association rules mining framework. OmicsARules embedded a new rule-interestingness measure, Lamda3, to evaluate the associated pattern and prioritize the most biologically meaningful gene associations.As demonstrated with DNA methlylation and RNA-seq datasets from breast invasive carcinoma (BRCA), esophageal carcinoma (ESCA) and lung adenocarcinoma (LUAD), Lamda3 achieved better biological significance over other rule-ranking measures. Furthermore, OmicsARules can illustrate the mechanistic connections between methlylation and transcription, based on combined omics dataset. OmicsARules is available as a free and open-source R package.ConclusionsOmicsARules searches for concurrent patterns among frequently altered genes, thus provides a new dimension for exploring single or multiple omics data across sequencing platforms.

Highlights

The improvements of high throughput technologies have produced large amounts of multi-omics experiments datasets
The improvements of high throughput technologies have enabled them to be precisely characterized at epigenetic, genomic, transcriptomic, proteomic and metabolomic levels [1,2,3]. While this opens the door to a systemsbased research approach, there is an urgent demand of novel methods to better illustrate the underlying mechanistic connections within or across different omics datasets
Our OmicsARules package, which embedded with a new rule-interestingness measure Lamda3, can evaluate the association rules to identify biologically significant patterns

Summary

Introduction

The improvements of high throughput technologies have produced large amounts of multi-omics experiments datasets Initial analysis of these data has revealed many concurrent gene alterations within single dataset or/and among multiple omics datasets. The improvements of high throughput technologies have enabled them to be precisely characterized at epigenetic, genomic, transcriptomic, proteomic and metabolomic levels [1,2,3]. While this opens the door to a systemsbased research approach, there is an urgent demand of novel methods to better illustrate the underlying mechanistic connections within or across different omics datasets. Our OmicsARules package, which embedded with a new rule-interestingness measure Lamda, can evaluate the association rules to identify biologically significant patterns

Methods

Results

Discussion

Conclusion