Question

Nonspecific filtering of microarray data

0

Entering edit mode

n.lu • 0

@nlu-13144

Last seen 5.2 years ago

Hi,

I am quite new to this field. Does anybody have recomendations on the criteria to use for prefiltering genes in genefilter package? I have found a few of them (row SDS, IQR, intensity based filtering) but am not sure which one/what combination to use. Or should I not prefilter at all?

microarray affymetrix microarrays pre-filtering filtering nonspecific filtering • 2.1k views

ADD COMMENT • link updated 7.5 years ago by Wolfgang Huber ★ 13k • written 7.5 years ago by n.lu • 0

0

Entering edit mode

It depends on what kind of analysis it is you're doing.

ADD REPLY • link 7.5 years ago b.nota ▴ 370

0

Entering edit mode

My experiment has 4 groups - 2 different genotypes- wild type and a knockout - and then in both a group with disease and a control (WT control, WT disease, knockout disease, knockout control). I want to do analysis of differentaly expressed genes between these groups, and than add some pathway analysis when I have the list of DE genes to see if they belong to a certain bio process/pathway..

ADD REPLY • link 7.5 years ago n.lu • 0

0

Entering edit mode

I don't think you'll need nonspecific filtering. I think you're better of using limma, it looks like a multi-factorial (two-way anova) design. Examples are in limma's user's guide for these kind of designs.

ADD REPLY • link 7.5 years ago b.nota ▴ 370

0

Entering edit mode

Thank you!

ADD REPLY • link 7.5 years ago n.lu • 0

0

Entering edit mode

Why do you think that? "need" is a strong term, but filtering (or weighting, see my reply below) can improve an analysis by giving better power.

ADD REPLY • link 7.5 years ago Wolfgang Huber ★ 13k

score 1 · Accepted Answer · 2017-06-02

1

Entering edit mode

Wolfgang Huber ★ 13k

@wolfgang-huber-3550

Last seen 12 weeks ago

EMBL European Molecular Biology Laborat…

In many cases, using data-driven weights for the multiple testing computation is a better approach than yes/no filtering, see e.g. the IHW vignette and the associated paper. The vignette also gives hints on the choice of the weighting covariate and diagnostic plots.

ADD COMMENT • link 7.5 years ago Wolfgang Huber ★ 13k