HI friends,
Is it possible to analysis the differential expression genes/transcripts form RNA-seq without any replications!!!. Actually I am handling the year old plant transcriptome data of one control and one treatment for each 8 differential conditions(assays) to find DEGs using the Tuxedo Suite (TopHat-Cufflinks-CummeRbund) and new Tuxedo Package suite (HISAT2-Stingtie-Ballgown). But resulted in very low number of genes/transcripts shows >+2 and
Hello Ellango,
You can consider transcripts with fold change more than 1 having p-value < 0.05. If you are having reference genome then you should use Tuxedo pipeline to extract differentially expressed transcripts but if you had denovo assembly then you should go with edgeR for DGE. Now the next point to be considered while using DGE is what dispersion values have you considered while doing pairwise differential gene expression. If you have samples in replicates then then software will calculate the dispersion value else you need to assign a dispersion value ranging between 0.1-1. In this case, while using dispersion value 0.1 you may got more differentially expressed transcripts with p-value
Without replicates, you cannot estimate which genes are differentially expressed using EdgeR or DESeq2. You can only calculate fold changes based on normalized read counts (preferentially CPM normalized by TMM method included in any EdgeR analysis) and apply a stringent fold-change cut-off to determine which genes are more or less expressed depending on the condition.
Hello,
NoiSeq, may be useful as it can simulate replicates and apply non-parametric test. http://www.bioconductor.org/packages/release/bioc/html/NOISeq.html.
There maybe other similar tools which may help you.
Good luck
Thanks Romit, Ashwani, Shaojun for your valuable suggestion...
I am accepting with Gautier, Actually I tried with DESeq, but it not accepting without replicate. Now I will try out the Ashwani suggestion "NOISeq" package for DEG analysis.
Hi Ellango,
Depending on the conditions tested and what questions you're asking from the data, it might be able to assume certain samples as biological replicates. I recommend you check your samples' clustering using a PCA plot (explained in the DESeq2 manual/workflow), this is a good way of exploring your data. For example, if you have 4 control samples and 4 treatment samples it might be that all your control samples make one o group and they together differentiate from your treatment samples. If this is the case, one could argue that for the purpose of analysing DEG between treatment and control one could consider all the control samples as replicates, same would apply for the treatment samples. However, the relevance of this approach will depend on the nature of the experimental setup.
In any case, handle with caution!
Good luck,
Anna
A little update for this post, DESeq2 begins to discontinue the option to run exploratory analysis without replicates in the 1.20.0 version, as stated in the update change-log:
________________________
o Beginning the deprecation of exploratory analysis of designs without replicates. Analysis of designs without replicates will be removed in the Oct 2018 release: DESeq2 v1.22.0, after which DESeq2 will give an error.
________________________
Source: https://bioconductor.org/packages/release/bioc/news/DESeq2/NEWS
I hope you had already solved the issue DE analysis with a non-replicated sample. If not, check LPEseq, focusing on that particular issue. Article LPEseq: Local-Pooled-Error Test for RNA Sequencing Experimen...
When in doubt, contact me via [email protected]
I am in the same case, I used an MSD plot to see the relationship between my libraries and from my control libraries I chose the ones that behaved like replicas due to their proximity. It must be taken into account that the conclusions will not have firm weight until another type of analysis is made
- Badges
- Science topic
- Similar topics
- Filing