[BioC] beadarray, bash, hulk, excludeBrihtBeadNeighbors, et al -- how to use practically?

[Mark Dunning]

Hi Anand,

Many thanks for your interest in our work on spatial artefacts. Our
current belief is to do things in the following order

1) Detect the phenomena described in the Smith et al (2010) paper. You
do need to run generateNeighbours before many of these functions. The
number of beads detected at this stage should be relatively small, and
would probably be removed by BASH anyway. However, their presence in
the BASH calculation might hide other defects that should be removed.

2) BASH to mask artefacts

3) HULK to smooth gradients

The order in which one does BASH and HULK could be a topic for
research. In cases of severe gradients across the array, you might get
a lot of beads masked at one edge of the array. However, if HULK were
run first these beads might be saved.

I guess you've already found the setWeights function in beadarray that
can save the weights found by BASH to the bead-level object? These
weights are multiplied by bead intensities when the bead summary data
are created. Currently we tend to use 0 or 1 as weights but there is
no reason why you couldn't use other values. A feature of the new
version of beadarray (version 2.0 and newer) is that you can save
weights under different names in the bead-level object. It should be
possible to do a series of tests for aberrant beads, save the results
separately and then combine in some way to produce the weights used in
summarisation.

Could you give more information about the errors you have with
EBImage? Which version of EBImage are you running? We tried the latest
of EBImage with the example data for the Smith et al paper and it
worked fine.

Regards,

Mark


On Sat, Oct 30, 2010 at 12:15 AM, Anand Patel <acpatel at gmail.com> wrote:
> OK.  After having used beadarray to get very good results from
> bead-level data, I've recently read all of
> the work on the multiple artifact removal methods worked out by the
> Cambridge group (refs at end of email).
>
> Indeed, I may have read them a bit too many times.  My question is
> based on figure 5 from the spatial
> phenomena paper -- the workflow:
> http://www.biomedcentral.com/content/supplementary/1471-
> 2105-11-208-s5.pdf -- link to the paper =
> http://www.biomedcentral.com/1471-2105/11/208 .
>
> As of this moment, I have several questions:
> 1.  In which sequence should one peform the steps in the workflow with
> respect to BASH and HULK?
> 2.  "While BASH removes beads from the analysis, HULK adjusts the
> values of the beads left after
> BASHing, and combination of the two tools requires care." -- how
> should one combine these tools with
> the material in the spatial phenomena paper?
> 3.  Does one run generateNeighbours before running the supplemental scripts
> "excludeBrightBeadNeighbours", "nearNonDecodedClusters", and
> "deviationFromGrid"?
> 4.  How does one set weights for all of those abnormal occurrences?
> There's no worked example using
> everything, just individual demonstrations.
>
> There's an amazing amount of work (as cited above), and I've already
> used HULK successfully for several
> analyses, but the new functions -- they're a bit rough.  EBImage's
> readImage throws up all kinds of flags
> about the TIFFs -- tags out of order, imagelength missing, etc.
> They're valid images though, Preview
> and Photoshop both open them correctly . . .
>
> Help is appreciated.
>
> Thanks,
> Anand
>
> Refs:
> Smith ML, Lynch AG. BeadDataPackR: A Tool to Facilitate the Sharing of
> Raw Data from Illumina BeadArray Studies. Cancer Informatics 2010;
> 2010:217-27.
>
> Smith ML, Dunning MJ, Tavaré S, Lynch AG. Identification and
> correction of previously unreported spatial phenomena using raw
> Illumina BeadArray data. BMC Bioinformatics 2010; 11:208.
>
> Lynch AG, Smith ML, Dunning MJ, Cairns JM, Barbosa-Morais NL, and
> Tavare S (2009). beadarray, BASH and HULK - tools to increase the
> value of Illumina BeadArray experiments. In A. Gusnanto, K.V. Mardia,
> & C.J. Fallaize (eds), Statistical Tools for Challenges in
> Bioinformatics, pp.33-37. Leeds, Leeds University Press.
>
> Cairns JM, Dunning MJ, Ritchie ME, Russell R, Lynch AG. BASH: A tool
> for managing BeadArray spatial artefacts Bioinformatics Oct 25 Epub
> ahead of print [pubmed]
>
> Dunning MJ, Barbosa-Morais NL, Lynch AG, Tavaré S, Ritchie ME.
> Statistical issues in the analysis of Illumina data. BMC
> Bioinformatics 2008 9:85
>
> Dunning MJ, Thorne NP, Tavaré S. Quality control and low-level
> statistical analysis of Illumina BeadArrays. REVSTAT - Statistical
> Journal, Volume 4, Number 1, March 2006, pages 1-30
>
> Anand C. Patel, MD, MS Bioinformatics
> Assistant Professor of Pediatrics and Medicine
> Division of Allergy/Pulmonary Medicine
> Department of Pediatrics
> Washington University School of Medicine
> 660 South Euclid Ave, Campus Box 8052
> St. Louis, MO 63110
> acpatel at wustl.edu
> acpatel at gmail.com
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>