Quality Control of Microarrays and the Importance of Human Computer Interaction

1
Quality Control of Microarrays and the Importance
of Human Computer Interaction

• Angela Burr
• Bioinformatics Capstone Project
• July 2004

2
Capstone Project

• Using computing solutions to improve the
  processing of data and analysis of results within
  the Drosophila Genomics Resource Center (DGRC)
  project of the Center for Genomics and
  Bioinformatics.
• Quality control issues will be addressed
  specifically.
• Focused on the aspects of human computer
  interaction between the biologists and the suite
  of computer programs I created to facilitate
  their research.

3
Background Information Microarrays

• Microarrays are biological experiments aimed to
  analyze gene expression of an organism.
• The DGRC produces Drosophila microarrays and
  completes hybridization experiments.

http//www.bioteach.ubc.ca/MolecularBiology/microa
rray/
4
Importance of Human Computer Interaction Component

• The large volume of data and various sources of
  data within microarray experiments make
  computational approaches essential.
• Continuous interaction develops between
  biological experiments and computational
  components.
• The computer experience of biologist is also a
  consideration.

5
Microarray Experiment
Microarray Production
Annotation
DNA QC
Hybridization QC
Computational Scripts
6
Objective of Quality Control in Microarray
Experiment

• Because of the expense of a microarray
  experiment, quality control (QC) is very
  important to ensure the integrity of the samples
  and the experiment.
• DNA QC
• Experimental Errors
• Version of genome
• Hybridization QC
• Experimental Errors

7
Microarray Experiment
Microarray Production
Annotation
DNA QC
Hybridization QC
Computational Scripts
8
Annotation Processing

• The primers for the DGRC project were developed
  by Incyte Genomics and used in previous
  microarray experiments. An annotation is
  available for these on NCBIs GEO (accession
  gpl20).
• The annoation contains information like the gene
  names, identification in biological databases,
  genome location and size of the DNA products.
• A script was used to convert the annotation into
  the appropriate format for the layout of the DGRC
  arrays.

9
Microarray Experiment
Microarray Production
Annotation
DNA QC
Hybridization QC
Computational Scripts
10
Processing of DNA Quality Control

• The image analysis data is in a convoluted order.
• The image analysis data is combined and
  reorganized to a single file by excel macros.
  Samples within the file are ordered by sample
  identification.
• Another macro identifies (flags) unreliable
  samples for lack of PCR product, multiple PCR
  products, low mass of PCR product, or incorrect
  molecular weight (size) of PCR product using the
  output of the previous macros and the annotation
  output.

11
DNA Quality Control Interfaces
12
DNA Quality Control Interfaces
13
DNA QC by Sample
14
DNA QC by Plate
15
Microarray Experiment
Microarray Production
Annotation
DNA QC
Hybridization QC
Computational Scripts
16
Processing Hybridization Quality Control

• Documentation for easier use of Bioconductor
• Scripts for easier importation
• Convert the necessary files into the correct
  format for Bioconductor
• Microarray layout file
• Microarray results file
• Add information about the control type and micro
  titer plate number for each sample
• based on a user-defined controls, quality control
  information, or the annotation file
• Script to create a file containing various
  exploratory plots for a hybridization data set
  for evaluation purposes
• For use before and after normalization

17
Hybridization QC
18
Usability Test

• An HCI (Human Computer Interaction) usability
  test was conducted to evaluate biologists
  interaction with some of the programs and lead to
  helpful suggestion. It is vital that biologist
  can easily use the programs.
• Recommendations were made for the system based on
  the test, which identified problems within the
  system.
• Modifications corrected the programs based on the
  recommendation from the HCI usability test.

19
Usability Test Problems and Solutions

• Main Identified Problems
• Help Files
• Problem Users were hesitant about using the
  help file, and the help file was lengthy.
• Resolution
• Reliance on the help file was reduced by adding
  more information to the interface.
• Context-sensitive help would be ideal, but may be
  difficult to implement. I am currently working
  on this problem.
• Naming conventions example
• Problem The names of the three programs were
  confusing, including information about the order
  in which they must be run. Users dont seem to
  know which file to select once they get to the
  menu.
• Resolution The files were renamed which
  included numbering to better distinguish the
  programs.

20
Usability Test Problems and Solutions cont.

• Main Identified Problems cont.
• Consistency throughout the applications example
• Problem The plate number was in a different
  format in two of the programs.
• Resolution The plate number was changed to the
  same numbering system.
• Programming error
• Problem A program crashed if a user browsed for
  a folder and then decided to cancel the
  operation.
• Resolution Error was fixed.
• Explanation of output
• Problem Users were unsure of what some of the
  fields in the resulting excel file (the final
  output) represented.
• Resolution A READ-ME worksheet was added to
  provide an explanation for the output,
  specifically a key for the fields of the
  worksheets.

21
Usability Test Conclusions

• In a post-test questionnaire, the evaluators were
  ask to rate their interaction with the
  application.
• This included questions on the usefulness, ease
  of use, and willingness to use the application.
• The application received excellent reviews from
  the users. This indicates that the users had a
  good feeling about the system and that it is in
  fact user-friendly.

22
Conclusions

• My capstone project helped to complete DNA and
  hybridization quality control for the DGRC
  microarray experiments.
• Within microarray experiments, data was processed
  more efficiently with the help of computational
  tools.
• The interaction between the biologist and the
  computational tool was key to the success of the
  tool.
• Developing thorough / useful documentation is
  also an important aspect of the interaction
  between the biologist and computational tools.

23
Future Work

• The DNA Quality Control methods offered a
  temporary solution. LIMS will offer a better,
  more reliable solution.
• Extension of project to include a complete
  solution for microarray data analysis which is an
  immediately need of the DGRC who are generating
  many experiments, but have yet to define a good
  method for analysis.
• A systematic flow of information from the raw
  microarray data to the completion of data
  analysis would be ideal.

24
References

• Bioconductor, www.bioconductor.org
• Center for Genomics and Bioinformatics, Indiana
  University, unpublished data, 2003-2004.
• DGRC, http//dgrc.cgb.indiana.edu
• Gene Expression Omnibus, http//www.ncbi.nlm.nih.g
  ov/geo, geo accession gpl20
• Recommendations for the Microarray Quality
  Control System, Alla Genkina and Stacey Sutton,
  2004.
• The R Project for Statistical Computing,
  www.r-project.org

25
Acknowledgements

• CGB genomics lab (Justen Andrews)
• Sun Kim, Bioinformatics advisor
• HCI involvement Youn Lim, Alla Genkina, and
  Stacey Sutton
• HCI Usability participants Kevin Bogart,
  Elizabeth Bohuski, Karmon Jones, Jacqeline Lopez,
  and Takuma Tsukahara.