Hey All
I want to use limma to identify differentially expressed genes. This is sample explanation.
This is my data frame
A B C D
hsa-miR-199a-3p, hsa-miR-199b-3p NA 13.13892 5.533703 25.67405
hsa-miR-365a-3p, hsa-miR-365b-3p 15.70536 52.86558 18.467540 223.51424
hsa-miR-3689a-5p, hsa-miR-3689b-5p NA 21.41597 5.964772 NA
hsa-miR-3689b-3p, hsa-miR-3689c 9.58696 44.56490 10.102051 13.26785
hsa-miR-4520a-5p, hsa-miR-4520b-5p 18.06865 28.06991 NA NA
hsa-miR-516b-3p, hsa-miR-516a-3p NA 10.77471 8.039662 NA
E
hsa-miR-199a-3p, hsa-miR-199b-3p NA
hsa-miR-365a-3p, hsa-miR-365b-3p 31.93503
hsa-miR-3689a-5p, hsa-miR-3689b-5p 24.26073
hsa-miR-3689b-3p, hsa-miR-3689c NA
hsa-miR-4520a-5p, hsa-miR-4520b-5p NA
hsa-miR-516b-3p, hsa-miR-516a-3p NA
Question# 1:
Should I do pairwise comparison between A&B A&C A&D A&E?? same for B&C B&D B&E??
Question# 2:
For Pairwise comparison I have to make a design and contrast matrix for that which I will use in lm function of limma. Can Any body can please guide me???
Suppose I have 1700 miRNAs in my data frame..
Best
Adnan Javed
•
written
by
adnanjaved1988 •
limma can do multi comparison. just need to make your design and contrast matrix.
yes I am confused how to make it?? Can you give me one example for my data ???
Try edgeR, it handles samples without replicates.
Note: But in my opinion, if you do not have replicates simply look for fold changes and decided the cutoff based on the distribution of fold change. There is also available tool which ranks based on generalized fold-change: http://www.ncbi.nlm.nih.gov/pubmed/22923299
Or if you have time-series data, look into significant profile/pattern of expression along time points. I recommend you to use STEM.
It sounds like you actually 2 have replicates, a treated, and untreated. If that's the case, this R code will get you started:
library(limma)
experiment = data.frame(treatment=rep(c("treated", "untreated"), 5), grp=rep(c("A", "B", "C", "D", "E"), each=2))
print(experiment)
design = model.matrix(~ 0 + treatment + grp, experiment)
print(design)
ctr = makeContrasts("treatmenttreated - treatmentuntreated", levels=design)
The output of that is:
treatment grp
1 treated A
2 untreated A
3 treated B
4 untreated B
5 treated C
6 untreated C
7 treated D
8 untreated D
9 treated E
10 untreated E
treatmenttreated treatmentuntreated grpB grpC grpD grpE
1 1 0 0 0 0 0
2 0 1 0 0 0 0
3 1 0 1 0 0 0
4 0 1 1 0 0 0
5 1 0 0 1 0 0
6 0 1 0 1 0 0
7 1 0 0 0 1 0
8 0 1 0 0 1 0
9 1 0 0 0 0 1
10 0 1 0 0 0 1
hey thanks Denver
But I don't have any biological replicates. Like If I would have then for Sample A I should have two or three sample with different conditions same for B C D E. All of these samples have been analysed only once.
For example for biological replicate (E) I would have two or three samples (same combination of drugs but with different environmental conditions.. But this is not the case. I only have total 5 samples (CDE are treatments) B is resistant to drugs. A is parental.
I have to use Edge R. for Edge R How can I design my Matrix for that???
•
written
by
adnanjaved1988 •
Log in to answer this question.
It looks like you have no replicates. LIMMA can do nothing for you in that case, it does not make sense to infer statistics from single sample experiments.
Hey These are 5 samples as every sample file has same miRNAs but the expression was changed as treatment drugs was applied so I made a data frame just used miRNAs one time instead of repeating for every sample now I want t look for up and down regulated miRNAs while pairwise comparisons whether a particular miRNA in particular treatment sample is up or down regulated..
I am sorry, but I do not understand what you mean. Are you trying to say you have 5 replicates per condition (i.e. five measurements of hsa-miR-199a-3p from parental MCF-7 cells)?
Hey David No I dont have any Biological replicates in my data. What I am trying to say I have total five samples.
Every sample has same miRNAs (2019). So by applying treatment drugs, Expression values have been measured so now I want identify potential miRNAs which show significant change of expression in treatments with respect to A..
Edge R is good option to deal with micro array data without replicates.
So while using Edge R how to design matrix for control and treatment?
Should I do pairwise comparison of just compare A with rest of samples?
Before doing any analysis, I would suggest you read Advice on RNASeq analysis without biological replicates for differential gene expression. and http://www.bioconductor.org/packages/release/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf (p. 19). I would suggest option #1.
hey this is not RNA-seq data ..
The purpose of this project is to compare miRNA present in exosomes from various fractions. To achieve this goal, Toray 3D-Gene technology is used for miRNA profiling. the company used Toray 3D-Gene Human miRNA Oligo chip 4-plex, targeting 2019 miRNA (based on miRBase release 19)
The point of linking these discussions was not to give you the go-to strategy, it was to point you towards what to do when you do not have biological replicates. The tl;dr (in my opinion): Don't use statistics if you have less than 3 biological replicates per treatment.
Saying you can use statistics even with one point is, again in my own opinion, the same as saying you can draw a straight line from one point.
David Sorry I really didn't get what you want to say. As I explain I don't have biological replicates so there is a Package (Edge R) which is use to handle micro array data without replicates. So Can you help me how I can perform analysis with that?? like contrast and design matrix construction ?