Clustering Gene Expression For Time Course Rna-Seq Data

Question

[Abhi]

Hi Guys

I am trying to read more about methods available/recommended for clustering gene expression data. Basically the input is a matrix of gene expression values from RNA-Seq differential expression analysis. In this case we have 10 different time points/ conditions.

We are trying to find out clusters which represent the states of the set of the genes across these different time points. Google/paper search did lead me to few hits but I was wondering if I can reach out to folks who have done this analysis in the past and learn from their experience.

Thanks! -Abhi

Answer 1

[Abhi]

Hey Guys

Just for completeness I found the following resources to be useful.

1. Concise review (somewhat old but useful) How does gene expression clustering work ? Patrik D'heaseleer : link

2. Data Clustering : A review by Jain, Murty And Flynn : link

3. Specific to gene expression clustering : Cluster Analysis for Gene Expression Data: A Survey :link

4. Judging the Quality of Gene Expression-Based Clustering Methods Using Gene Annotation : link

5. Tool: The C Clustering Library : link

6. Estimating the number of clusters in a data set via the gap statistic : link basically a method to determine an optimum #K(partitions) for partition based clustering

7. Problems in gene clustering based on gene expression data. Journal of Multivariate Analysis 90, 44–66.

8. Clustering tools available in R ( somewhat exhaustive list as far as the basic methods are concerned) link

9. A roadmap of clustering algorithms: finding a match for a biomedical applicationlink

-Abhi

[Istvan Albert]

thanks for the followup - nice link collection!

Answer 2

[Steve Lianoglou]

I'm not sure if you are looking for something different than "normal" clustering? I mean, does the fact that this is timecourse data influence the clustering approach you want to take?

Anyway, I think Section 7 of the DESeq vignette could be helpful, as they first shoot their data through a variance stabilized transformation before doing exploratory analysis (heatmaps, PCA, etc). which (by the looks of what they present there) seem like a good thing to try.

[Ryan Thompson]

Yes, the best bet for now is probably to use DESeq's variance-stabilizing transformation and then use tools designed for microarray analysis.

Answer 3

[Alex Reynolds]

Sean Thomas developed a procedure called rank expectation that he used to cluster chromatin accessibility in areas of the Drosophila genome at five stages of development (see Fig. 4). The test procedure is generally applicable to situations where you want to discern groups or clusters of data by signal, like gene expression, chromatin accessibility, etc.

Answer 4

[boczniak767]

For time-course microarray expression data (8 time points) I've used successfully mfuzz (details)

which performs soft clustering.

Answer 5

[Houkto]

HI there,

I use BioLayout Express3D which has MCL clustering tool and you can visualize the results LINKS

Answer 6

[sabaghianamir70]

you need too set your data in 10 time point like this way 0-1 0-2 0-3 0-4 0-5 ... you can do it even with deseq2 and get output for other analysis