Microarray Data Analysis Pipelines - User-Level Workflow Design: A Bioinformatics Perspective - page 125

• DilutionAssessmentBasic assesses a single preprocessing method based

on its application to the dilution benchmark data set.

•

DilutionAssessmentComparative assesses and compares two preprocess-
ing methods based on their application to the dilution benchmark data
set.

•

SpikeInAssessmentBasic assesses a single preprocessing method based
on its application to a spike-in benchmark data set.

•

SpikeInAssessmentComparative assesses and compares two preprocess-
ing methods based on their application to a spike-in benchmark data set.

•

SpikeInAssessment2Basic assesses a single preprocessing method based
on its application to a spike-in benchmark data set using an alternative
assessment function.

•

SpikeInAssessment2Comparative assesses and compares two preprocess-
ing methods based on their application to a spike-in benchmark data set
using an alternative assessment function.

Note that the SIBs in this library call R scripts that bundle particular
Bioconductor functionality, following the approach that has already been
taken for the case study on LC/MS data preprocessing and analysis work-
flows based on the xcms Bioconductor package [158]. That is, the methods
provided by the underlying libraries are not directly passed to the user level,
but manually curated and wrapped into domain-specific accessible units of
functionality. In his diploma thesis [334], Clemens von Musil developed a
plugin for the automatic creation of SIBs from R packages based on their
description files (which are similar to the manual pages of Unix operating
systems). The plugin, called jR , accomplishes its task pretty well, and its
use was demonstrated by means of different examples (matrix multiplication,
lexical coverage, microarray data classification). However, the granularity of
the provided functionality and thus of the workflows remains close to the
corresponding R code, so that workflow design with jR-generated SIBs is too
low-level for a framework that advocates working with coarse-grained func-
tionality at the user level. Therefore jR has not been used for the integration
of Bioconductor functionality into Bio-jETI.

6.2.2 Basic Microarray Data Analysis Pipeline

A basic, more or less minimal, example of a microarray analysis pipeline built
with the SIB library described above is shown in Figure 6.6: The HGU95A
spike-in benchmark data set is loaded and then preprocessed and filtered
using the corresponding AffyExpress SIBs. After a differential expression
analysis, the PubMed abstracts of articles potentially related to the top 10
differentially expressed genes are retrieved and saved. In this form, the anal-
ysis workflow corresponds clearly to the abstract microarray data analysis
pipeline of Figure 6.5.