Transcript assembly, differential expression, and differential regulation for RNA-Seq


Site Map

News and updates

    New releases and related tools will be announced through the mailing list

Getting Help

    Questions about Cufflinks and Cuffdiff should be posted on our Google Group. Please use for private communications only. Please do not email technical questions to Cufflinks contributors directly.


Related Tools

  • Monocle: Single-cell RNA-Seq analysis
  • CummeRbund: Visualization of RNA-Seq differential analysis
  • TopHat: Alignment of short RNA-Seq reads
  • Bowtie: Ultrafast short read alignment




Getting started

Install quick-start

Installing a pre-compiled binary release

In order to make it easy to install Cufflinks, we provide a few binary packages to save users from occasionally frustrating process of building Cufflinks, which requires that you install the Boost libraries. To use the binary packages, simply download the appropriate one for your machine, untar it, and make sure the cufflinks,cuffdiff and cuffcompare binaries are in a directory in your PATH environment variable.

Building Cufflinks from source

In order to build Cufflinks, you must have the Boost C++ libraries (version 1.47 or higher) installed on your system. See below for instructions on installing Boost.

Installing Boost

  1. Download Boost and the bjam build engine.
  2. Unpack bjam and add it to your PATH.
  3. Unpack the Boost tarball and cd to the Boost source directory. This directory is called the BOOST_ROOT in some Boost installation instructions.
  4. Build Boost. Note that you can specify where to put Boost with the --prefix option. The default Boost installation directory is /usr/local. Take note of the boost installation directory, because you will need to tell the Cufflinks installer where to find Boost later on.

    If you are on Mac OS X, type (all on one line):

    bjam --prefix=<YOUR_BOOST_INSTALL_DIRECTORY> --toolset=darwin architecture=x86 address_model=32_64 link=static runtime-link=static --layout=versioned stage install
  5. If you are on a 32-bit Linux system, type (all on one line):

    bjam --prefix=<YOUR_BOOST_INSTALL_DIRECTORY> --toolset=gcc architecture=x86 address_model=32 link=static runtime-link=static stage install

    If you are on a 64-bit Linux system, type (all on one line):

    bjam --prefix=<YOUR_BOOST_INSTALL_DIRECTORY> --toolset=gcc architecture=x86 address_model=64 link=static runtime-link=static stage install

Installing the SAM tools

  1. Download the SAM tools
  2. Unpack the SAM tools tarball and cd to the SAM tools source directory.
  3. Build the SAM tools by typing make at the command line.
  4. Choose a directory into which you wish to copy the SAM tools binary, the included library libbam.a, and the library headers. A common choice is /usr/local/.
  5. Copy libbam.a to the lib/ directory in the folder you've chosen above (e.g. /usr/local/lib/)
  6. Create a directory called "bam" in the include/ directory (e.g. /usr/local/include/bam)
  7. Copy the headers (files ending in .h) to the include/bam directory you've created above (e.g. /usr/local/include/bam)
  8. Copy the samtools binary to some directory in your PATH.

Installing the Eigen libraries

  1. Download Eigen
  2. Unpack the Eigen tarball and cd to the Eigen source directory.
  3. Copy the Eigen/ subdirectory someplace on your system where you keep header files (e.g. /usr/local/include)

Building Cufflinks

  1. Unpack the Cufflinks source tarball:
    tar zxvf cufflinks-0.7.0.tar.gz
  2. Change to the Cufflinks directory:
    cd cufflinks-0.7.0
  3. Configure Cufflinks. If Boost is installed somewhere other than /usr/local, you will need to tell the installer where to find it using the --with-boost option. Specify where to install Cufflinks using the --prefix option.
    ./configure --prefix=/path/to/cufflinks/install --with-boost=/path/to/boost --with-eigen=/path/to/eigen
    If you see any errors during configuration, verify that you are using Boost version 1.47 or higher, and that the directory you specified via --with-boost contains the boost header files and libraries. See the Boost Getting started page for more details. If you copied the SAM tools binaries to someplace other than /usr/local/, you may need to supply the --with-bam configuration option.
  4. Finally, make and install Cufflinks.
    make install

Testing the installation

  1. Download the test data
  2. In the directory where you placed the test file, type:
    cufflinks ./test_data.sam

    You should see the following output:
    [bam_header_read] EOF marker is absent. The input is probably truncated.
    [bam_header_read] invalid BAM binary header (this is not a BAM file).
    File ./test_data.sam doesn't appear to be a valid BAM file, trying SAM...
    [13:23:15] Inspecting reads and determining fragment length distribution.
    > Processed 1 loci.                            [*************************] 100%
    Warning: Using default Gaussian distribution due to insufficient paired-end reads in open ranges.  
    It is recommended that correct paramaters (--frag-len-mean and --frag-len-std-dev) be provided.
    > Map Properties:
    >       Total Map Mass: 102.50
    >       Read Type: 75bp x 75bp
    >       Fragment Length Distribution: Truncated Gaussian (default)
    >                     Estimated Mean: 200
    >                  Estimated Std Dev: 80
    [13:23:15] Assembling transcripts and estimating abundances.
    > Processed 1 loci.                            [*************************] 100%
  3. Verify that the file transcripts.gtf is in the current directory and looks like this (your file will have GTF attributes, omitted here for clarity)
    test_chromosome Cufflinks       exon    53      250     1000    +       . 
    test_chromosome Cufflinks       exon    351     400     1000    +       . 
    test_chromosome Cufflinks       exon    501     550     1000    +       .

Common uses of the Cufflinks package

Cufflinks includes a number of tools for analyzing RNA-Seq experiments. Some of these tools can be run on their own, while others are pieces of a larger workflow. The complexity of your workflow depends on what you want to achieve with your analysis. For a complete discussion of how Cufflinks can help you with your analysis, please see our protocol paper. The paper includes a diagram (Figure 2) describing how the various parts of the Cufflinks package (and its companion tool TopHat) fit together. As of version 2.2.0, you can also run Cuffquant and Cuffnorm to make large scale analyses easier to handle. The figure below is an updated version of Figure 2 showing how the two utilities released after the protocol paper appeared fit into the workflow:

You can use Cuffquant to pre-compute gene expression levels for each of your samples, which can save time if you have to re-run part of your analysis. Using Cuffquant also makes it easier to spread the load of computation for lots of samples across multiple computers. If you don't want to perform differential expression analysis, you can run Cuffnorm instead of Cuffdiff. Cuffnorm produces simple tables of expression values that you can look at in R (for example) to cluster samples and perform other follow up analysis.

Discovering novel genes and transcripts

RNA-Seq is a powerful technology for gene and splice variant discovery. You can use Cufflinks to help annotate a new genome or find new genes and splice isoforms of known genes in even well-annotated genomes. Annotating genomes is a complex and difficult process, but we outline a basic workflow that should get you started here. The workflow also excludes examples of the commands you'd run to implement each step in the workflow. Suppose we have RNA-Seq reads from human liver, brain, and heart.

  1. Map the reads for each tissue to the reference genome
  2. We recommend that you use TopHat to map your reads to the reference genome. For this example, we'll assume you have paired-end RNA-Seq data. You can map reads as follows:

    tophat -r 50 -o tophat_brain /seqdata/indexes/hg19 brain_1.fq brain_2.fq tophat -r 50 -o tophat_liver /seqdata/indexes/hg19 liver_1.fq liver_2.fq tophat -r 50 -o tophat_heart /seqdata/indexes/hg19 heart_1.fq heart_2.fq

    The commands above are just examples of how to map reads with TopHat. Please see the TopHat manual for more details on RNA-Seq read mapping.

  3. Run Cufflinks on each mapping file
  4. The next step is to assemble each tissue sample independently using Cufflinks. Assemble each tissue like so:

    cufflinks -o cufflinks_brain tophat_brain/accepted_hits.bam
    cufflinks -o cufflinks_liver tophat_liver/accepted_hits.bam
    cufflinks -o cufflinks_heart tophat_liver/accepted_hits.bam
  5. Merge the resulting assemblies
  6. assemblies.txt:
    Now run the merge script:
    cuffmerge -s /seqdata/fastafiles/hg19/hg19.fa assemblies.txt

    The final, merged annotation will be in the file merged_asm/merged.gtf. At this point, you can use your favorite browser to explore the structure of your genes, or feed this file into downstream informatic analyses, such as a search for orthologs in other organisms. You can also explore your samples with Cuffdiff and identify genes that are significantly differentially expressed between the three conditions. See the workflows below for more details on how to do this.

  7. (optional) Compare the merged assembly with known or annotated genes
  8. If you want to discover new genes in a genome that has been annotated, you can use cuffcompare to sort out what is new in your assembly from what is already known. Run cuffcompare like this:
    cuffcompare -s /seqdata/fastafiles/hg19/hg19.fa -r known_annotation.gtf merged_asm/merged.gtf
    Cuffcompare will produce a number of output files that you can parse to select novel genes and isoforms.

Identifying differentially expressed and regulated genes

There are two workflows you can choose from when looking for differentially expressed and regulated genes using the Cufflinks package. The first workflow is simpler and is a good choice when you aren't looking for novel genes and transcripts. This workflow requires that you not only have a reference genome, but also a reference gene annotation in GFF format (GFF3 or GTF2 formats are accepted, see details here). The second workflow, which includes steps to discover new genes and new splice variants of known genes, is more complex and requires more computing power. The second workflow can use and augment a reference gene annotation GFF if one is available.

Differential analysis without gene and transcript discovery

  1. Map the reads for each condition to the reference genome
  2. We recommend that you use TopHat to map your reads to the reference genome. For this example, we'll assume you have paired-end RNA-Seq data. Suppose you have RNA-Seq from a knockdown experiment where you have two biological replicates of a mock condition as a control and two replicates of your knockdown.

    Note: Cuffdiff will work much better if you map your replicates independently, rather than pooling the replicates from one condition into a single set of reads.

    Note: While an GTF of known transcripts is not strictly required at this stage, providing one will improve alignment sensitivity, and ultimately, the accuracy of Cuffdiff's analysis.

    You can map reads as follows:

    tophat -r 50 -G annotation.gtf -o tophat_mock_rep1 /seqdata/indexes/hg19 \
      mock_rep1_1.fq mock_rep1_2.fq
    tophat -r 50 -G annotation.gtf -o tophat_mock_rep2 /seqdata/indexes/hg19 \
      mock_rep2_1.fq mock_rep2_2.fq
    tophat -r 50 -G annotation.gtf -o tophat_knockdown_rep1 /seqdata/indexes/hg19 \
      knockdown_rep1_1.fq knockdown_rep1_2.fq
    tophat -r 50 -G annotation.gtf -o tophat_knockdown_rep2 /seqdata/indexes/hg19 \
      knockdown_rep2_1.fq knockdown_rep2_2.fq
  3. Run Cuffdiff
  4. Take the annotated transcripts for your genome (as GFF or GTF) and provide them to cuffdiff along with the BAM files from TopHat for each replicate:
    cuffdiff annotation.gtf mock_rep1.bam,mock_rep2.bam \

Differential analysis with gene and transcript discovery

  1. Complete steps 1-3 in "Discovering novel genes and transcripts", above
  2. Follow the protocol for gene and transcript discovery listed above. Be sure to provide TopHat and the assembly merging script with an reference annotation if one is available for your organism, to ensure the highest possible quality of differential expression analysis.

  3. Run Cuffdiff
  4. Take the merged assembly from produced in step 3 of the discovery protocol and provide it to cuffdiff along with the BAM files from TopHat:
    cuffdiff merged_asm/merged.gtf liver1.bam,liver2.bam brain1.bam,brain2.bam
    As shown above, replicate BAM files for each conditions must be given as a comma separated list. If you put spaces between replicate files instead of commas, cuffdiff will treat them as independent conditions.