Clustering of genes using FPKM

Question

[Ron]

Hi,

I understand that bioconductor packages such as edgeR and DESeq require readcounts for further downstream analysis.I have FPKM values from cufflinks.Can I just use the values to produce a heatmap?

Also,I want to get the top expressed genes from normals vs Tumors.Any suggestions.

Best,

Ron

Answer 1

[GouthamAtla]

If you have bam files, run cuffdiff. The cuffdiff output will have differentiallly expressed isoforms, genes, FPKM values for genes and isoforms etc.

The cuffdiff output can be read in to cummeRbund package which has many visualisations including heatmaps on FPKM values.

If you want to use edgeR or DESeq, use htseq-count to get the raw counts per gene/exon and feed that data into edger/DESeq.

Answer 2

[Hamid Ghaedi]

Old question but this may be of help for those dealing with the same issue:

For clustering, you need to transform your data into a z-score. There is a Bioconductor package, zFPKM do this on FPKM reads.

[ATpoint]

The purpose of zFPKM is to define a meaningful thresholds for active genes. For clustering a simple scale() is sufficient.

[Rob]

Hi Hamid I also want to cluster my genes for plotting a heatmap. How should I cluster then I have upregulated genes together and downregulated together? I want to see patterns and difference in patterns between two groups (treatment and control)

[Hamid Ghaedi]

@rhasanvandj,

Do a differential expression analysis, Set a cut-off to retrieve DE genes, Use a package for heatmap plotting. Please check this tuturial from Kevin Blighe.

[jared.andrews07]

Please open your own question rather than piggybacking on another if you have one. This helps us keep the site organized and allows other to find your question more easily.