Meeting Monday 24th February

[adam....@ucdconnect.ie]

Hi all,

I'm wondering if there's an appetite for a meeting on Monday week (24th February), as we haven't had one since before Christmas!

I think it would be a good idea to have a structure to the meeting, so we can set an agenda beforehand of things we'd like to get through.

Does anyone have stuff they're currently working on that they'd like to chat about? Or any papers/topics you've read that might be of interest?

Adam

[Karsten Hokamp]

Hi Adam,

I'd be up for it! I've got a 5-minute presentation on a normalisation issue that I had meant to present ages ago. It's about a case where DESeq can produce very wrong values. Will be happy to show that unless we have enough other topics.

Another thing I'd be interested in is gathering a list of next-gen-related meetings/conferences that might be worth attending this year.

Karsten

--
You received this message because you are subscribed to the Google Groups "OpenSequencing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to opensequencin...@googlegroups.com.
Visit this group at http://groups.google.com/group/opensequencing.
For more options, visit https://groups.google.com/groups/opt_out.

[Adam Dinan]

Hi Karsten,

OK great, so we'll go ahead with the meeting next Monday in the interactive hub at 2pm - if you can give a talk on the DESeq issue that would be nice.

I'll also put down the conferences issue to be discussed. I have a couple of issues that I'd like to bring up also, one being the use of TPMs vs RPKMs for normalising gene expression, which I know was mentioned at a meeting before.

If anyone has something they'd like to discuss then feel free to post and we can talk about it there.

--

Adam Dinan

PhD student

School of Medicine & Medical Science,

Conway Institute,

University College Dublin,

Belfield, Dublin 4, Ireland.

Tel: +353833143372

Email: adam....@ucdconnect.ie

[Markus Schroeder]

There are several papers showing different kinds of biases when using RPKM normalisation. Mainly an over-correction of short exons and to favour transcripts with long exons as DE genes, resulting in a smaller p-value. There are also reported differences between samples, sequencing runs and data sets when using RPKMs:

[HTML] RNA sequencing reveals two major classes of gene expression levels in metazoan cells

D Hebenstreit, M Fang, M Gu… - Molecular systems …, 2011 - Wiley Online Library

... The fragments used to estimate intron and intergenic RPKM were based on randomizations using
the ... The accuracy of RNA-seq is biased toward longer and more highly expressed genes, eg ... To
explore how this accuracy bias affects the shape of the LE distribution, we studied ...

[HTML] Bias detection and correction in RNA-Sequencing data

W Zheng, LM Chung, H Zhao - BMC bioinformatics, 2011 - biomedcentral.com

... expression levels from RNA-Seq data, such as reads per kilobase of gene length per million
reads (RPKM), are biased in terms ... Compared to previously proposed base level correction
methods, our method reduces bias in gene-level expression estimates more effectively. ...

[HTML] Reducing bias in RNA sequencing data: a novel approach to compute counts

F Finotello, E Lavezzo, L Bianco, L Barzon… - BMC …, 2013 - biomedcentral.com

... Exon length bias and GC-content effect (Jiang, "cell"). ... Counts or RPKMs are computed using
totcounts, maxcounts, RPKM-corrected totcounts (RPKM) and totcounts corrected with
within-lane full-quantile normalization over exon length (FullQ), and averaged across libraries. ...

Best,
Markus

[Adam Dinan]

Thanks for sharing Markus,

I have noticed that many papers, even in relatively high impact journals, are continuing to use the RPKM metric to normalise their data and it doesn't seem to have an issue getting through peer review. Though I wonder if this is largely because of the fact that RPKMs were described in a Nature methods paper while the TPM concept was introduced in a less well known journal.

Anyway, if anyone else has issues they'd like to address then please post as we should have time to discuss other stuff.