New release of RobiNA - a new tool providing a complete RNA-Seq based DE analysis wor

[marnilush]

We are happy to announce a new and improved release of our RNA-Seq
analysis tool RobiNA (see http://seqanswers.com/wiki/RobiNA and http://mapman.gabipd.org/web/guest/robin).
We included several new features ( e.g. GC-content bias normalization,
import of precomputed counts tables). The new version is based on the
R-2.15.0 and Bioconductor 2.10.

Installer packages for windows and Mac OS X include an embedded R engine -
Users working on linux can install the "lightweight" package and use an
R installation on their system (has to be R-2.15.0, though). RobiNA will
check whether all required R packages are installed and try to download
missing packages (requires either admin permissions or an R configuration
that installs new packages in a user-writable directory).

We invite all interested users to give RobiNA a try and are very grateful
for any feed back that can help us to make the tool more useful and
stable - so thanks in advance for bug reports and feature requests!

[happyfly]

Hello marnilush,

I found RobiNA is interesting to try. I have been trying with the RobiNA with 6 SAM files (1 control, 1 treatment, 3 biological reps each). The program has been running for 2 days with 50% CPU since data were loaded. I am in the step for DESeq/EdgeR analysis but can't click to design my experiment in this window. I am wondering if this is memory problem. Do you have more detail recommendations for memory requirement than the one you have in the Robin/RobiNA manual? I have 4GB RAM and 32-bit Windows. I have been searching if somebody has been trying this program like me but haven't found any.

[marnilush]

hi happyfly and thanks for your feedback. When started on windows machines, RobiNA should automatically grab 80% of the free RAM as the maximal java heap space. In case that was not enough (maybe that happened on your computer since 2 days @ 50% CPU sounds like the machine was very busily using the hard disk - maybe it was
swapping heavily) there might have been a memory exception in the background. Did RobiNA display a memory warning message? If it did there most probably was a memory exception possibly leading to inconsistent behaviour downstream in the analysis workflow.

It would be great if you could give me more details about the problem you experienced on the experiment design panel - You wrote you couldn't click to design the experiment. Does that mean it was impossible to draw arrows between the boxes representing the treatments? Or was the whole GUI completely unresponsive? In case you were unable to draw arrows, please try holding down different combinations of the CTRL and SHIFT (and maybe also ALT) keys while click-dragging between the boxes. Depending on your keyboard/language settings and OS version the keys you need to press might differ.

I hope this helped a bit,
cheers,
Marc

[happyfly]

Marc,

Thank you very much for the prompt response.

I don't think I have seen a memory warning message.

I tried twice with the same set of data.

On the first time, I could draw the arrow when the data were loading. But by accidence, the program was closed.

I started over. This time, I left the computer for some hours for data loading. When I came back, the window of Design experiment was like I have described, the mouse pointer became "<--->" whenever it was in this window area. The "idle" appeared in the bottom right corner. In Window Task Manager the program is still "running", not "non-responding".

I will start again one more time to see what will happen.

[marnilush]

Hi and thanks again for your feedback. The behaviour you describe could
be explained by an exception that was thrown in the background - when
checking the task manager, did RobiNA still produce some CPU load? I am
not completely sure what you mean by the mouse pointer became "<--->" - did it show a double arrow when hovering over the RobiNA window?

It would be very helpful if you could tell me exactly which version of
Windows in combination with which version of Java you are using so i
could try reproducing your environment an a machine here. If you like you could also try running RobiNA from a command line window - this would allow you/me to see whether any exceptions are thrown. To do that, simply click the "run" dialog in the windows start menu and type "cmd". Now navigate to the folder in which you installed RobiNA (e.g. C:\Program Files\RobiNA) using the "cd" command and then start RobiNA by typing:

java -jar -Xmx3G Robin2.exe

the program should start up and you should see messages appear in the console window. In addition to this, there should be a file named sth like
"YOUR_EXPERIMENT_NAME.log" in the log subfolder - Most of the exceptions that are thrown while RobiNA runs should be logged there as well.

Best greetings and thanks for your patience.
cheers,
Marc

[marnilush]

Hi again - i think i might have found the reason for the crash you are
reporting: The BAM/SAM importer needs BAM/SAM files with a valid header section. In some cases, since it is optional according to the SAM format specification (http://samtools.sourceforge.net/SAM1.pdf), the header might not be included in the files.

If this is the case you should see an error message when starting RobiNA from the command line as i have described in the last post - and I'd be grateful if you could send me the error message text for investigation (e.g. to [email protected]).

cheers,
Marc

[happyfly]

The followings are our dicussion via email: Italic font is by Marc

The mouse pointer became "<--->" - it is a double arrow when hovering over the RobiNA window. At this, javaw took ~50% load of CPU.

I did start RobiNA one more time and my computer was in "Loading data". This time, there is almost no CPU (1-2%) load. After waiting for ~5-6 hours, when it was still Loading data, I clicked on the Next and the program was closed. This is similar to the first time I tried.

I also ran the command line and the results were:
Error occurred during initialization of VM
Could not reserve enough space for object heap
Could not create the Java virtual machine.

Our computer is Windows Version 6.1.7601.

I see that your windows version is quite up to date (Windows 7 SP1). However,
the behaviour you are describing really sounds to me like there was an error
when RobiNA tried to read the SAM files you supplied - Did you have a chance
to check the files and make sure they have a complete header section?

The error you got when starting RobiNA from the command line indicates that the
Java virtual machine could not allocate the amount of memory that was asked for.
Maybe try starting java with a different memory setting; e.g. by typing

java -jar -Xmx2048M Robin2.exe

this command requests 2048 MB (2 GB) of memory for the Java heap -
if this still causes the same error try lowering the maximal heap space
even more until the machine is happy.


2048 MB (2 GB) of memory for the Java heap didn't work but 1.5 GB did. There was also an error message on parsing SAM file (was copied on the txt file attached). Is this the header error like you said? If yes, I will try to find SAM samples with header and test the program again. Interestingly, this time, when the data is loading, I click on next, the program showed up a notice about the error and closed (it didn't if Robina was started by the clicking on the RobiNA icon on the Desktop).

Assume that the SAM files work, I am wondering whether my computer with 1.5GB of memory can handle big sam files (e.g: wildtype + mutant, 2 reps/each; ~8GB total).

concerning RobiNA, the error message really indicates the error i suspected
to be the reason for the crash: The SAM file seems to be lacking the header
section. If you do also have BAM files (this is essentially the binary version
of SAM) of the same data you should use the samtools to generate SAM
files with a complete header section and then try to load these into RobiNA.
I do not think that the amount of memory you have available on your machine
is limiting - the log file does also not report any memory related problems.
The problem really only was the SAM file format and bigger files in the correct
format should work fine - they are actually not loaded into memory completely
but are read line by line so that memory consumption is kept small.


Run by command lines:
C:\Users\nhamngoc>cd "C:\Program Files\RobiNA"
C:\Program Files\RobiNA>java -jar -Xmx1572M Robin2.exe
The “estimate RPKM expression values” didn’t work for me. The function square was faint and I was not able to tick on it.
It took 2mins to load 2 SAM files (5.75GB). Unfortunately, I don’t have samples with biological replicates. When the files had been loaded, there was an notice “Your data doesn’t include replicates... “ and the program closed.

Run by RobiNA icon on Windows
The “estimate RPKM expression values” didn’t work either. It did the work but generate non-sense graph (the reason might be only one rep I have)

I will try later when I find a data set with >=2 reps.

[marnilush]

Hi happyfly,

when importing BAM/SAM files, the RPKM estimation function is
deactivated because to estimate RPKM values, RobiNA needs the
gene (or transcript) lengths. These are automatically recorded
when doing the alignments within RobiNA.

For upcoming releases i will evaluate the possibility to extract the
length information from the BAM/SAM header and then enable
RKPM estimation for BAM/SAM import as well. Please note that
for the estimation of differential expression, estimation of RPKM
values is not necessary and the values are not used.

Concerning the biological replicate problem: Which analysis package
did you use? DESeq or edgeR?

cheers,
Marc

[steindorff]

Problems with build a genome reference

Hi,

I'm running the RobiNA pipeline with my RNA-seq data. So that's my problem:

When I use a Transcriptome reference (fasta with all transcripts), the index build work well. But when I use a genome reference, both of them (gtf and gff3) got an error. In gtf format, the index builder don't accept the 9° column (features) and with the gff3 the message "null gene appears in more than one chromossome: scaffold_1 and scaffold_10" appears.

I already run the tophat pipeline with these same files and it worked well. What should I do to solve these "reference" problems?

Thanks!

[sg45653]

New user of RobiNA - need help

Hello,

I am a new user of RobiNA (many thanks for developing such a tool BTW) and am trying to analyze a paired-end RNASeq dataset of 3 samples (6 fastq files). I reached upto the stage of downloading the mouse reference genome but when downloading got an error message of "RobiNA is running out of memory and may become unstable". About 30 seconds after this, the tool disappeared.

My computer is a Dell Precision T3600 workstation with 8GB RAM and an Intel Xeon processor (CPU E5-1620, 3.6GHz).

Any help is greatly appreciated.

Sujoy

[marnilush]

Depending on the operating system you are running RobiNA on, the amount
of memory that is allocated to RobiNA might differ. Under windows memory
allocation is dynamic and will use 75% of the free RAM at start time. In case
you have many other programs open, that might not be sufficient. You could
try closing all other applications and then restarting RobiNA. Theoretically
your machine should have enough RAM to run the analysis.

In case that does not help, you could start RobiNA from a command console
and allocate more memory by issuing the command

java -jar -Xmx7G Robin2.jar

(or .exe in case of windows)

That would reserve maximally 7GB for RobiNA.

Another point is the paired-end data - the current version of RobiNA
cannot work with paired-end data. I am working on this and it will be
possible to import paired-end reads with the next release. Currently, you can
analyse paired-end data in RobiNA if you run the alignments externally (e.g.
using bowtie/bowtie2 or any other aligner of you choice than can produce BAM/SAM
output) and import the BAM/SAM files.

I hope this helps,
bests,
Marc

[rain bow]

using RobinA by giving external counts table as input

Hi there,
I have RNA-seq data without replicates and since I m not so much aware of statistical parameters, I am using RobinA for analysing fold changes in different samples. I have already got counts table generated by using CLC platform and now I need to use Deseq to get comaparable fold changes among 5 different samples. I've used various parameters, still got no useful result.
Please guide me how to use it in a simplified way.

Regards,

[marnilush]

Hi rain bow,

thanks for your message. In principle, analysing data without replication
should be possible (although it is not very recommendable because the
results won't be very reliable - You might want to have a look at this
post http://seqanswers.com/forums/showpos...04&postcount=2
for more details on why biological replication is needed). It would help if you could give me more
details on how the analysis did not yield useful results. Did the program
crash or are the results you got simply not meeting your expectations?

In case the program crashed there should be a ".log" file in the log subfolder
of your project directory - it would be great if you could send me this file
because it contains details on why the crash might have happened.
(simply send it to lohse_at_mpimp-golm.mpg.de)

cheers,
Marc

[Jin72]

DESeq in RobiNA

Hi there,

I have been using RobiNA for my RNA-Seq data set. By using the DESeq for the duplicate samples between treatment group and control group, I have got so small number of DEGs (< 10). Here is my detail condition for the run.
P-value :0.05, FDR: BH, log2fold =1: activated, dispersion=pooled.
It looked too stringent. How could I get more DEGs? FYI, by RPKM method with the condition (p-value :0.05, FDR:0.001), more than 300 of DEGs were obtained.

Thank you in advance for your reply.

Jin-hyoung