Comparison between two outputs of limmas toptable function from filtered and unfiltered dataset
Entering edit mode
Last seen 7.6 years ago

After searching about filtering method for DEG analysis, finally I decide to to use panp package for my work,I exclude those prob-IDs that present in less than 10% of each groups(control vs basal like breast cancer), although in 100 prob-IDs that came from taptable function in limma,there are a lot of common genes(with variation in position in ranked list) in two list that come from toptable on 1- unfilteret and 2- filtered dataset, but after validation of these to gene list with survival analysis on an independent dataset, I found that my filtered data gives me a better output , but its p-values seems insignificant


output from unfiltered dataset test:

                              P.Value                adj.P.Val              B

213706_at           4.95E-49               1.10E-44               100.5409

204388_s_at       1.98E-48               2.21E-44               99.17014

43427_at              5.43E-48               4.03E-44               98.17258

221928_at           9.04E-48               5.04E-44               97.66887

201890_at           2.02E-47               9.02E-44               96.87277

204997_at           4.50E-47               1.67E-43               96.08317

207092_at           1.36E-46               4.31E-43               94.99363

205913_at           2.05E-46               5.70E-43               94.58617

49452_at              2.58E-46               6.39E-43               94.35622

204389_at           2.01E-45               4.48E-42               92.32877

218039_at           3.72E-45               7.53E-42               91.72103

206030_at           1.46E-44               2.70E-41               90.37272

212741_at           1.89E-44               3.24E-41               90.11262

208383_s_at       5.34E-44               8.50E-41               89.08835


output from filtered dataset


design <- model.matrix(~factor(filtereddata$Disease))

fit1 <- lmFit(filtereddata,design)

ebayes1 <- eBayes(fit1)

tab1 <- topTable(ebayes1, coef=2, adjust="fdr", n=150)


                              P.Value                adj.P.Val              B

221928_at           0.097476              0.999917              -4.29479

204570_at           0.10648                0.999917              -4.35075

49452_at              0.116574              0.999917              -4.40763

212741_at           0.116969              0.999917              -4.40974

205913_at           0.124306              0.999917              -4.44763

210298_x_at      0.128623              0.999917              -4.46879

213071_at           0.128718              0.999917              -4.46925

221747_at           0.128972              0.999917              -4.47046

216331_at           0.135798              0.999917              -4.50225

205382_s_at       0.13713               0.999917              -4.50825

221748_s_at       0.137664              0.999917              -4.51064

203548_s_at       0.140671              0.999917              -4.52388

now I want to know your opinion about this results??

thanks in advance

limma taptable DEG • 1.3k views
Entering edit mode
Last seen 14 hours ago
WEHI, Melbourne, Australia

From your brief description, the filtering you have done appears to be incorrect because you seem to have assessed expression separately in the two groups of samples. You absolutely must not use knowledge of which sample belongs to which group when you do the filtering.

Perhaps there are other problems as well, because the results look a bit strange to me.

Entering edit mode

After I merge two eset with COMBAT method in insilicomerging package, my merged eset has 30 control ,60 basal like breast cancer and 60 luminal A in the next step I subset the merge eset into 3 esets that each one is include control, basal like and luminalA(because I thought it's better i.e. to filter non-informative gene in each group than performing a blind filtering on all group)

here is my codes ,maybe in the next step I should not use COMBAT method for merging my esets







selcontrol <- rowSums(mypcallscontrol=="P")>3

filteredcontrol <- controleset[selcontrol ,]

###########after I did the same codes for basal group,again I merged them to get an eset with 2 filtered group(control vs basal like) 



design1 <- model.matrix(~factor(finaleset$Disease))

fit1 <- lmFit(finaleset, design1)

ebayes1 <- eBayes(fit1)

tab <- topTable(ebayes1, coef=2, adjust="fdr", n=50)

Thank you very much


Login before adding your answer.

Traffic: 714 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6