Please Help To Understand The Square Root Of Jensen-Shannon Divergence

Question: Please Help To Understand The Square Root Of Jensen-Shannon Divergence

6.2 years ago by

Du, Jianguang • 380 wrote:

Dear All, I am looking for the differential splicing events between cell types. However the Cuffdiff gives output using "the square root of Jensen- shannon divergence" to measure the difference. Although I tried my best to understand "the definition of the square root of Jensen-shannon divergence", I still could not understand the meaning of a specific value of "the square root of Jensen-shannon divergence". I would appreciate it very much if anyone let me know how to covert "the square root of Jensen-shannon divergence" into "fold". For example, how much "the square root of Jensen-shannon divergence" is "2 fold difference" equal to. Thanks in advance. Jianguang

rna-seq cuffdiff • 2.8k views

ADD COMMENT • link •

modified 6.2 years ago by Jennifer Hillman Jackson ♦ 25k • written 6.2 years ago by Du, Jianguang • 380

6.2 years ago by

Jennifer Hillman Jackson ♦ 25k

United States

Jennifer Hillman Jackson ♦ 25k wrote:

Hello Jianguang, Fold is included in the Cuffdiff output. Section "Differential expression tests", first file, column #9. http://cufflinks.cbcb.umd.edu/manual.html Hopefully this helps, Jen Galaxy team -- Jennifer Jackson http://galaxyproject.org

ADD COMMENT • link written 6.2 years ago by Jennifer Hillman Jackson ♦ 25k

Hi Jen, Thank you for your answer. However, the output file "transcript differential expression testing" gives the ratio (log2 of the fold change) of FPKM of a specific transcript between two conditions, which means this fold change in FPKM does not take the overall gene expression into consideration (the expression of one gene many be much higher in condition A than condidtion B) and therefore can not be used as difference of alternative splicing. What I am doing is looking for the difference of splicing between two cell types by examining the output file "splicing differential expression testing". In this file, column 10 gives the value of "sqrt(JS)" (the explaination of it is "The splice overloading of the primary transcript, as measured by the square root of the Jensen- Shannon divergence computed on the relative abundances of the splice variants", and the value is never larger than 1). My understanding is that, this value has already taken the overall gene expression into consideration. But I do not know how much "sqrt(JS)" equals to 2 fold of change because I want to focus the alternative splicing with 2 fold difference between two cell types. Do you know how to convert the value of "sqrt(JS)" into "fold"? In addition, how to understand the sentence "The splice overloading of the primary transcript"? If one gene have 3 transcript: A, B and C and the expression of these transcripts is: A=60%, B=25%, and C=15%, do you mean the primary transcript is A? Does the Cuffdiff take the overall expression (A+B+C=100%) or just the primary transcript (A=60%) into consideration when calculates the ratio of transcript B? Actually, it would be much easier for us if Cuffdiff calculates the ratio of the expression of the alternatively spliced exon to overall gene expression, and then compare between conditions. Thanks in advance, Jianguang ________________________________________ To: Du, Jianguang Cc: galaxy-user@lists.bx.psu.edu; closeticket@galaxyproject.org Subject: Re: [galaxy-user] Please help to understand the square root of Jensen-Shannon divergence Hello Jianguang, Fold is included in the Cuffdiff output. Section "Differential expression tests", first file, column #9. http://cufflinks.cbcb.umd.edu/manual.html Hopefully this helps, Jen Galaxy team -- Jennifer Jackson http://galaxyproject.org

ADD REPLY • link written 6.2 years ago by Du, Jianguang • 380

Hello Jianguang, The tool package CummeRbund would likely be of interest http://compbio.mit.edu/cummeRbund/ Support can be obtained through the these mailing lists: http://lists.sourceforge.net/lists/listinfo/bowtie-bio-announce They would also be best folks to address your new questions about the the underlying mechanism of how splice overload is calculated. In general, converting to fold starts by converting a data metric into a log value (log2 generally). A google can bring up many resources that describe the calculation and various tools (R, etc.). This is common for expression analysis, but whether this is recommended by the tool authors for Cuffdiff output analysis is another good question to ask them. Good luck with your project, Jen Galaxy team -- Jennifer Jackson http://galaxyproject.org

ADD REPLY • link written 6.2 years ago by Jennifer Hillman Jackson ♦ 25k

Similar posts • Search »