Advice : Building Transcript Models Using Cufflinks

Question

[Abhi]

Hi Guys

Keeping it short I want to know which is a better approach for building transcripts models using cufflinks.

If I have an older set of gene models (as a .gtf file) for a species and some new RNA-Seq data, which approach is recommended:

1. Use RABT (reference annotation based transcript) assembly by providing the gene models to Cufflinks, or
2. Build the transcriptome de-novo using Cufflinks and then use Cuffmerge to merge the new model with the older one.

Our goal is to improve the existing gene models using new RNA-Seq data.

Thanks!

-Abhi

[Arun]

I have to mention that Augustus with RNA-Seq data and reference as inputs resulted in better gene models for us than Cufflinks. We tested it with Arabidopsis thaliana Col-0. So maybe its also nice to compare gene models from Augustus and Cufflinks with Cuffcompare? Just a thought.

Answer 1

[Malachi Griffith]

Either of those approaches may prove useful. I would recommend that you try both and compare the outcome.

Here are some threads with advice you might want to consider first:

Introduction

• RNA-seq pre-processing before de novo transcriptome assembly
• Is there any reason to do de novo transcript assembly if a reference is available?
• Does it makes sense to assemble all my RNA-Seq samples as though they were one?
• de novo small RNA transcriptome assembly for non-model species?

Tool recommendation

• What programs are there for transcriptome assembly?
• Cufflinks transcript assembly
• Trinity package

Interpretation, post-processing, and applications

• How are RNAseq Transcripts Assigned?
• Annotating a novel transcriptome assembly
• De novo transcriptome and comparative genomics
• Downstream analysis of assembled transcripts (cufflinks, trinity)
• Determining FPKM of de novo assembly of Transcriptome