Partek 6.5 beta and RNA-seq

[mbjohnson]

Hello all,

I have just been trying out the Partek Genomics Suite beta 6.5 release which includes an RNA-seq workflow. I have successfully imported and processed Illumina reads that were already aligned using ELAND (the s_1_sorted.txt file). However, in trying to import raw reads (s_1_sequence.txt) in fastq format and performing the alignment from within Partek using bowtie, I run into problems. The import appears to work fine, and I then see a summary of the number of aligned reads per file/lane. But when I try to then map the reads to RefSeq or AceView genes, I get all zeros in the raw and RPKM read counts columns.

Is anyone else using the Partek RNA-seq workflow with Illumina fastq output? Any ideas where the issue might lie?

Thanks,
Matt

[RNA-seq]

I guess you did't have the alias file during import. One quick way to check is try the "plot chromosome view" in the workflow and check the chromosome name from the dropdown ont the top of the right panel. If you see weird chromosome names like Chr GI then that is the problem. You need the alias file to convert these chr GIs to regular chromosome name so that your reads can be mapped to Aceview. Try to contact [email protected] for the alias file, or download the newest version of 6.5beta and re-import your sample and generate the alias file.

[mbjohnson]

Thanks for the tip. I can see now that even though I have the alias file, it must be the wrong one. None of the chromosome names in my bowtie output match the ones in the alias file... Is it possible this is a genome build problem, hg18 vs hg19 mix-up?

I'm in touch with Partek support and will let you all know what I learn.

[mbjohnson]

Solved: I was using the "contigs" instead of the "assembly" version of the bowtie index. For anyone else trying out the Partek RNA-seq workflow, make sure you use the hg18 assembly index, "h_sapiens_asm.ebwt.zip" from the bowtie site.

[capricy]

Anybody knows what statistical distribution Partek Genomics Suite is fitting to get the differential expression?

[mbblack]

Anybody knows what statistical distribution Partek Genomics Suite is fitting to get the differential expression?

For RPKM values, I'm pretty sure it just log2 transforms them and then treats them as normal, since ANOVA is their default method for differential analysis.

I'd note too, since this was an old thread, Partek's current release is GS 6.6 with the latest update on July 13, 2012. Much of the most recent changes have been with next gen sequence features. For example, loading mapped BAM files and indexing to annotation goes much, much faster now than in GS 6.5 (and the windows 64-bit version is more stable with large next gen data sets then GS 6.5 was - I run GS 6.6 on an AMD 8-core Windows 7 machine with 32Gb RAM and 6.6 has improved performance and stability a lot).

[wxw]

Anybody knows what statistical distribution Partek Genomics Suite is fitting to get the differential expression?

In Partek Genomics Suite, RPKM is the default input to ANOVA in RNA-seq workflow. However, you have the option to choose which normalization method you want to use with your raw reads by changing the setting manually. There are dozens of alternative choices. There are also other choices for the statistical test besides ANOVA, and in Partek Flow, the best statistical test is automatically chosen using a Gene/Transcript-Specific Analysis which might be ANOVA, negative binomial, or Poisson - whichever best fits the gene.

[tharan]

Please tell me what is the best normalization method to use with my raw read

[wxw]

In Partek Flow, we provide several different options for normalization, as we don't in general believe in a "one size fits all" approach to analysis. For example, the number of samples in your study may be important in this decision.

A related issue is what distribution is assumed in a downstream statistical analysis. For that we analyze each gene (or transcript) individually to see if it is better modeled by a log-normal, Poisson, or negative binomial distribution.

If you are a current Partek user, you can contact us and we can advise you better based on your data.

[tharan]

Thanks for the clarification on normalization issues. I have pooled 3 samples in my miRNA sequencing work and getting in to the data analysis. I will get back to you with more queries. I have another problem, how to reduce clonality because in my last run I had 40% clonality. I am just following the work flow instructions by Life Tech and running the samples on Ion Torent.

Thanks