edgeR - 3-dimensional Multidimensional scaling

0

Entering edit mode

Paraskevas Iatropoulos ▴ 20

@paraskevas-iatropoulos-6265

Last seen 10.3 years ago

Hi! it would be of great help, if someone could suggest a solution to the following problem. I used edgeR to analyze RNA Seq expression data (3 different cell lineages with 2 biological replicates for each lineage). I used the plotMDS () function to get a 2D plot of multidimensional scaling. I was asked to perform a 3D plot but I am not able to find a solution According to Cell 151:559â575, where they report to perform 3D- MDS , and a web guide (Quick-R) I used the cmdscale () function as follows: d <- dist( mydata ) # Calculate the euclidean distances between the rows fit <- cmdscale (d, eig =TRUE, k=3) # k is the number of dimensions x <- fit$points[,1] y <- fit$points[,2] z <- fit$points[,3] Then I use the scatterplot3D software to obtain the grafic in 2D or 3D. However, the 2D plot obtained with cmdscale using the counts per million ( cpm ) is much different compared to that obtained with plotMDS . Moreover, the results with the above cmdscale procedure are strange from a biological point of view, so I suppose that I did something wrong. Since the dimensions in the edgeR graphic are "leading logFC ", I calculated the log2 fold-chage (log2 [ cpm of the replicate / average expression in all 6 samples]) and I get a 2D graphic very similar but not identical to edgeR . In this way I can also have a 3D plot. I am not sure if it is right to use the log2-FC to get. Do you think it is ok ? Is there a way to perform a 3D-MDS with edgeR ? Can you suggest me a way to perform the 3D- MDS , please? Thank you in advance. Paraskevas -- Paraskevas Iatropoulos, MD Geneticist Clinical Research Centre for Rare Diseases âALDO E CELE DACCO â IRCCS - Mario Negri Institute for Pharmacological Research Via G.B. Camozzi , 3 Ranica ( BG ) - Italy e-mail: iatropoulos @ marionegri .it ---------- L'IRCCS - Istituto di Ricerche Farmacologiche Mario Negri desidera ringraziare i 13.768 donatori, di cui non conosce i nomi, che nel 2011 hanno devoluto il loro 5 per mille a sostegno della nostra attivitÃ e assicurare loro che tutto quanto ricevuto sarÃ esclusivamente impiegato per le nostre ricerche. Siamo naturalmente molto grati a tutti coloro che anche quest'anno hanno deciso di rinnovare ancora la fiducia nel nostro lavoro con la destinazione del loro 5 per mille. A coloro che stanno per farlo ricordiamo di firmare ed indicare il nostro codice fiscale 03254210150 nel riquadro "Finanziamento agli enti della ricerca scientifica e dell'universitÃ . Per maggiori informazioni: IRCCS - Istituto di Ricerche Farmacologiche Mario Negri, Via La Masa 19 - 20156 Milano Tel: +39 02 390141 - Fax: +39 02 3546277 +39 02 39001918 Internet: www.marionegri.it, mnegri@marionegri.it [[alternative HTML version deleted]]

edgeR edgeR • 4.1k views

ADD COMMENT • link updated 11.1 years ago by Mark Robinson ▴ 880 • written 11.1 years ago by Paraskevas Iatropoulos ▴ 20

0

Entering edit mode

Mark Robinson ▴ 880

@mark-robinson-4908

Last seen 6.1 years ago

Dear Paraskevas, Maybe all you need is to set 'ndim=3' in your call to plotMDS() ? > However, the 2D plot obtained with cmdscale using the counts per million ( cpm ) is much different compared to that obtained with plotMDS . Have a careful read of: ?plotMDS.DGEList ?plotMDS In short, plotMDS() does something very different with a DGEList object than with a matrix of numbers. Then, maybe try something like: # Taken from the plotMDS.DGEList help page ngenes <- 1000 nlib <- 6 counts <- matrix(rnbinom(ngenes*nlib, size=1/10, mu=20),ngenes,nlib) rownames(counts) <- paste("Gene",1:ngenes) group <- gl(2,3,labels=c("Grp1","Grp2")) counts[1:200,group=="Grp2"] <- counts[1:200,group=="Grp2"] + 10 y <- DGEList(counts,group=group) y <- calcNormFactors(y) > mds <- plotMDS(y, top=200) > mds$cmdscale.out [,1] [,2] Sample1 0.6296663 2.11210712 Sample2 -3.2338291 -0.68439921 Sample3 2.1070395 0.06385855 Sample4 0.2781657 0.60249359 Sample5 0.9806977 -3.12230808 Sample6 -0.7617402 1.02824803 > mds <- plotMDS(y, top=200, ndim=3) > mds$cmdscale.out [,1] [,2] [,3] Sample1 0.6296663 2.11210712 -1.3976172 Sample2 -3.2338291 -0.68439921 1.2745170 Sample3 2.1070395 0.06385855 2.8492196 Sample4 0.2781657 0.60249359 -0.4326530 Sample5 0.9806977 -3.12230808 -1.5085139 Sample6 -0.7617402 1.02824803 -0.7849524 Note, you don't *need* to set the 'top' parameter. Then, just plot these in 3D. Hope that helps. Best, Mark ---------- Prof. Dr. Mark Robinson Bioinformatics, Institute of Molecular Life Sciences University of Zurich http://tiny.cc/mrobin On 27.11.2013, at 16:15, Paraskevas Iatropoulos <paraskevas.iatropoulos at="" marionegri.it=""> wrote: > > Hi! > > it would be of great help, if someone could suggest a solution to the following problem. I used edgeR to analyze RNA Seq expression data (3 different cell lineages with 2 biological replicates for each lineage). I used the plotMDS () function to get a 2D plot of multidimensional scaling. I was asked to perform a 3D plot but I am not able to find a solution > > According to Cell 151:559?575, where they report to perform 3D- MDS , and a web guide (Quick-R) I used the cmdscale () function as follows: > d <- dist( mydata ) # Calculate the euclidean distances between the rows > fit <- cmdscale (d, eig =TRUE, k=3) # k is the number of dimensions > x <- fit$points[,1] > y <- fit$points[,2] > z <- fit$points[,3] > Then I use the scatterplot3D software to obtain the grafic in 2D or 3D. > > However, the 2D plot obtained with cmdscale using the counts per million ( cpm ) is much different compared to that obtained with plotMDS . Moreover, the results with the above cmdscale procedure are strange from a biological point of view, so I suppose that I did something wrong. Since the dimensions in the edgeR graphic are "leading logFC ", I calculated the log2 fold-chage (log2 [ cpm of the replicate / average expression in all 6 samples]) and I get a 2D graphic very similar but not identical to edgeR . In this way I can also have a 3D plot. > > I am not sure if it is right to use the log2-FC to get. Do you think it is ok ? Is there a way to perform a 3D-MDS with edgeR ? Can you suggest me a way to perform the 3D- MDS , please? > > Thank you in advance. > > Paraskevas > > > > -- > Paraskevas Iatropoulos, MD > Geneticist > Clinical Research Centre for Rare Diseases ?ALDO E CELE DACCO ? > IRCCS - Mario Negri Institute for Pharmacological Research > Via G.B. Camozzi , 3 > Ranica ( BG ) - Italy > e-mail: iatropoulos @ marionegri .it > > ---------- > L'IRCCS - Istituto di Ricerche Farmacologiche Mario Negri desidera ringraziare i 13.768 donatori, di cui non conosce i nomi, che nel 2011 hanno devoluto il loro 5 per mille a sostegno della nostra attivit? e assicurare loro che tutto quanto ricevuto sar? esclusivamente impiegato per le nostre ricerche. Siamo naturalmente molto grati a tutti coloro che anche quest'anno hanno deciso di rinnovare ancora la fiducia nel nostro lavoro con la destinazione del loro 5 per mille. > > A coloro che stanno per farlo ricordiamo di firmare ed indicare il nostro codice fiscale > > 03254210150 > > nel riquadro "Finanziamento agli enti della ricerca scientifica e dell'universit?. > > Per maggiori informazioni: > > IRCCS - Istituto di Ricerche Farmacologiche Mario Negri, Via La Masa 19 - 20156 Milano > Tel: +39 02 390141 - Fax: +39 02 3546277 +39 02 39001918 > Internet: www.marionegri.it, mnegri at marionegri.it > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

ADD COMMENT • link 11.1 years ago Mark Robinson ▴ 880

0

Entering edit mode

Dear Mark, thanks for your help. I followed your instructions and it worked great! Paraskevas -- Paraskevas Iatropoulos, MD Geneticist Clinical Research Centre for Rare Diseases ?ALDO E CELE DACCO? IRCCS - Mario Negri Institute for Pharmacological Research Via G.B. Camozzi, 3 Ranica (BG) - Italy e-mail: iatropoulos at marionegri.it ----- Messaggio originale ----- Da: "Mark Robinson" robinson at imls.uzh.ch> A: "Paraskevas Iatropoulos" <paraskevas.iatropoulos at="" marionegri.it=""> Cc: bioconductor at r-project.org Inviato: Mercoled?, 27 novembre 2013 20:59:35 Oggetto: Re: [BioC] edgeR - 3-dimensional Multidimensional scaling Dear Paraskevas, Maybe all you need is to set 'ndim=3' in your call to plotMDS() ? > However, the 2D plot obtained with cmdscale using the counts per million ( cpm ) is much different compared to that obtained with plotMDS . Have a careful read of: ?plotMDS.DGEList ?plotMDS In short, plotMDS() does something very different with a DGEList object than with a matrix of numbers. Then, maybe try something like: # Taken from the plotMDS.DGEList help page ngenes <- 1000 nlib <- 6 counts <- matrix(rnbinom(ngenes*nlib, size=1/10, mu=20),ngenes,nlib) rownames(counts) <- paste("Gene",1:ngenes) group <- gl(2,3,labels=c("Grp1","Grp2")) counts[1:200,group=="Grp2"] <- counts[1:200,group=="Grp2"] + 10 y <- DGEList(counts,group=group) y <- calcNormFactors(y) > mds <- plotMDS(y, top=200) > mds$cmdscale.out [,1] [,2] Sample1 0.6296663 2.11210712 Sample2 -3.2338291 -0.68439921 Sample3 2.1070395 0.06385855 Sample4 0.2781657 0.60249359 Sample5 0.9806977 -3.12230808 Sample6 -0.7617402 1.02824803 > mds <- plotMDS(y, top=200, ndim=3) > mds$cmdscale.out [,1] [,2] [,3] Sample1 0.6296663 2.11210712 -1.3976172 Sample2 -3.2338291 -0.68439921 1.2745170 Sample3 2.1070395 0.06385855 2.8492196 Sample4 0.2781657 0.60249359 -0.4326530 Sample5 0.9806977 -3.12230808 -1.5085139 Sample6 -0.7617402 1.02824803 -0.7849524 Note, you don't *need* to set the 'top' parameter. Then, just plot these in 3D. Hope that helps. Best, Mark ---------- Prof. Dr. Mark Robinson Bioinformatics, Institute of Molecular Life Sciences University of Zurich http://tiny.cc/mrobin On 27.11.2013, at 16:15, Paraskevas Iatropoulos <paraskevas.iatropoulos at="" marionegri.it=""> wrote: > > Hi! > > it would be of great help, if someone could suggest a solution to the following problem. I used edgeR to analyze RNA Seq expression data (3 different cell lineages with 2 biological replicates for each lineage). I used the plotMDS () function to get a 2D plot of multidimensional scaling. I was asked to perform a 3D plot but I am not able to find a solution > > According to Cell 151:559?575, where they report to perform 3D- MDS , and a web guide (Quick-R) I used the cmdscale () function as follows: > d <- dist( mydata ) # Calculate the euclidean distances between the rows > fit <- cmdscale (d, eig =TRUE, k=3) # k is the number of dimensions > x <- fit$points[,1] > y <- fit$points[,2] > z <- fit$points[,3] > Then I use the scatterplot3D software to obtain the grafic in 2D or 3D. > > However, the 2D plot obtained with cmdscale using the counts per million ( cpm ) is much different compared to that obtained with plotMDS . Moreover, the results with the above cmdscale procedure are strange from a biological point of view, so I suppose that I did something wrong. Since the dimensions in the edgeR graphic are "leading logFC ", I calculated the log2 fold-chage (log2 [ cpm of the replicate / average expression in all 6 samples]) and I get a 2D graphic very similar but not identical to edgeR . In this way I can also have a 3D plot. > > I am not sure if it is right to use the log2-FC to get. Do you think it is ok ? Is there a way to perform a 3D-MDS with edgeR ? Can you suggest me a way to perform the 3D- MDS , please? > > Thank you in advance. > > Paraskevas > > > > -- > Paraskevas Iatropoulos, MD > Geneticist > Clinical Research Centre for Rare Diseases ?ALDO E CELE DACCO ? > IRCCS - Mario Negri Institute for Pharmacological Research > Via G.B. Camozzi , 3 > Ranica ( BG ) - Italy > e-mail: iatropoulos @ marionegri .it > > ---------- > L'IRCCS - Istituto di Ricerche Farmacologiche Mario Negri desidera ringraziare i 13.768 donatori, di cui non conosce i nomi, che nel 2011 hanno devoluto il loro 5 per mille a sostegno della nostra attivit? e assicurare loro che tutto quanto ricevuto sar? esclusivamente impiegato per le nostre ricerche. Siamo naturalmente molto grati a tutti coloro che anche quest'anno hanno deciso di rinnovare ancora la fiducia nel nostro lavoro con la destinazione del loro 5 per mille. > > A coloro che stanno per farlo ricordiamo di firmare ed indicare il nostro codice fiscale > > 03254210150 > > nel riquadro "Finanziamento agli enti della ricerca scientifica e dell'universit?. > > Per maggiori informazioni: > > IRCCS - Istituto di Ricerche Farmacologiche Mario Negri, Via La Masa 19 - 20156 Milano > Tel: +39 02 390141 - Fax: +39 02 3546277 +39 02 39001918 > Internet: www.marionegri.it, mnegri at marionegri.it > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor ---------- L'IRCCS - Istituto di Ricerche Farmacologiche Mario Negri desidera ringraziare i 13.768 donatori, di cui non conosce i nomi, che nel 2011 hanno devoluto il loro 5 per mille a sostegno della nostra attivit? e assicurare loro che tutto quanto ricevuto sar? esclusivamente impiegato per le nostre ricerche. Siamo naturalmente molto grati a tutti coloro che anche quest'anno hanno deciso di rinnovare ancora la fiducia nel nostro lavoro con la destinazione del loro 5 per mille. A coloro che stanno per farlo ricordiamo di firmare ed indicare il nostro codice fiscale 03254210150 nel riquadro "Finanziamento agli enti della ricerca scientifica e dell'universit?. Per maggiori informazioni: IRCCS - Istituto di Ricerche Farmacologiche Mario Negri, Via La Masa 19 - 20156 Milano Tel: +39 02 390141 - Fax: +39 02 3546277 +39 02 39001918 Internet: www.marionegri.it, mnegri at marionegri.it

ADD REPLY • link 11.1 years ago Paraskevas Iatropoulos ▴ 20

Login before adding your answer.