Automated RNA-Seq Analysis for Non-Model Organisms on the Cloud

Summary:

We present a cloud-based de novo transcriptome pipeline to analyze RNA-Seq data from non-model organisms.

Background:

RNA-Seq is a promising technique for understanding gene expression and metabolic networks. Non-model organisms often lack a reference genome, so de novo transcriptome assembly is the preferred strategy to study them. A number of open-source assemblers and downstream tools have been developed for this purpose. However, this is a complex process because of large data sizes, big demands for hardware memory and CPUs, and the complexity of benchmarking for effective and efficient use of the algorithms. Setting up these complex statistical analyses on local machines or clusters remains challenging.

To enable flexible and scalable data processing, we have built an end-to-end de novo transcriptome analysis pipeline on the cloud. This pipeline distributes the execution of computational sub-jobs over different machine types in order to optimize for the memory and CPU requirements of different algorithms. It was based on recommendations from the Trinity package and consists of the following major modules: Trinity Assembler, Bowtie Aligner, RSEM for gene expression quantification, EdgeR for differential expression (DE) analysis, and BlastX for functional annotations. We present a case study of this pipeline where we analyze test data and present the report with results statistics and metrics for performance and runtime. We also demonstrate its integration with the Tuxedo pipeline for improved performance strategy and a visualization feature for exploring DE results.

References:
Haas B, et al.: De novo transcript sequence reconstruction from RNA-seq using the Trinity platform for reference generation and analysis. Nature Protocols 2013, 8:1494–1512.