maxdLoad

1
maxdLoad StreptoBASE

• Tools for building a gene expression data
  resource
• OR
• How to find a Good Kebab Recipe

2
Thank You!

• Michael Wilson, Joe Wood, Sadeeh Maleki
• Vassilis Mersinias, Helena Kieser, John Bishop
• Workshop attendees

3
Outline

• MIAME, MAGE and kebabs
• maxd database system
• StreptoBASE

4
MIAME, MAGE and kebabs

• What is an object model and why bother going to
  the trouble of using one to store microarray data?

5
MIAME

• Minimum Information About a Microarray Experiment
• specification of necessary information to
  describe a microarray experiment
• concept rather than a format
• generic to all sorts of microarray experiments
• MAGE is an object model based on MIAME

6
What is an object model?

• Object Model
• formal description
• circuit diagrams
• data design tool
• databases
• XML
• programming variables
• a recipe

7
Different Recipes
Assemble on plate
Assemble on plate
Grill
Grill
Marination
8
Building a model structure

• determine what varies between recipes
• e.g., salad, bread, chicken, skewers, grilling,
  marination, assembly on plate etc.
• assign names to these things
• e.g., ingredients, utensils, cooking steps,
  assembly steps etc.
• break down into sub-categories
• e.g., Ingredients vegetables, meat, spices etc.

9
Common Features

• Name of recipe
• Ingredients
• chicken
• Utensils
• skewer

• Preparation steps
• chopping
• Cooking steps
• grilling

10
A Recipe model
11
A Recipe database
12
A Recipe XML

• ltRecipe nameChicken Kebab Supremegt
• ltIngredient nameChicken /gt
• ...
• ltPreparationMethod nameSpicy Marination
  IngredientSpicy Marinade amount200ggt Food
  item was soaked in Spicy Marinade for 3 hours.
  lt/PreparationMethodgt
• ...
• ltPreparedIngredient nameMarinated Chicken
  PreparationMethodSpicy Marination
  IngredientChicken /gt
• ...
• lt/Recipegt

13
An object model is...

• formal description
• translates directly to data structures, e.g., XML
• which software can be made to import, manipulate
  and export
• differences between recipes contrasted, and
  information stored
• more in common, easier to construct
• devil in detail

14
Why model?

• every good cook book has
• index at the back
• contents at the front
• pictures
• to help you find decide on recipes you want
• REASON 1
• Search

15
Why model?

• every good cook mixes ideas from different
  recipes
• REASON 2
• Compare

16
Why bother?

• Reading a cook book allows a human to Search and
  Compare ...
• ... so, for microarray data, why bother with a
  model?

17
Large datasets, many of them

• pubmed
• search limited to abstracts
• free-text mining of entire articles
• infancy
• need for ability to search experimental
  descriptions more effectively

18
Search and compare

• usage
• browsing and retrieval
• analysis

19
data-mining analysis

• classes assigned
• from experimental design parameters
• e.g., treatment time 0, time 1 etc., pooling
  replicates
• other possibly interesting classes
• operator, array type, array batch, scanner,
  harvesting protocol, cell strain, treatment
  compound
• normally lost or buried in theses (REASON 3)

20
other questions...

• I would like to
• see under what conditions does a list of genes
  co-express, and when they don't, across all
  experiments in a database
• find out if there are any experiments where
  metabolism is affected in the latter parts of the
  growth phase of my favourite organism
• find out if a batch of arrays is producing
  particularly bad results, or explains the results
  I see better than the classes I have assigned

21
your drug-treatment experiment

• after normalisation filtering
• genes that show little variation within
  replicates
• differentially expressed genes
• find most common GO-terms
• find other experiments that
• are a drug-treatment experiment
• have genes with the same GO-terms over-expressed
• in the same organism

22
why use an object model - summary

• this allows
• powerful search (REASON 1)
• comparisons between classes within and across
  experiments (REASON 2)
• storage of useful information normally lost or
  buried (REASON 3)
• data storage in a way that resembles the thing it
  is trying to describe
• aiding data-mining and analysis

23
Data capture Direct Benefits

• Force thought about experimental design
• Post-docs and students leave
• Data not fully exploited
• Unexplored avenues, comparisons with future
  results
• Lab books vary in quality
• Ability to export MAGE-ML
• e.g., for Array Express
• Analysis Support

24
Data Capture Indirect Benefits

• Data dissemination
• A transcriptome resource for everyone
• Numbers meaningless alone
• Resource stores the context
• e.g., reclassification by batch of arrays
• Potential for complex queries
• find all the genes upregulated under drug
  treatment x in experiment 204 from Jones lab,
  where transcription is not significantly affected
  in any of mutant in list y of experiments 14-58
  in Smiths lab, that have ontology terms related
  to metabolism and

25
maxd database system

• MIAME-compliant microarray data-capture and
  export tools

26
maxdLoad2 software

• allows
• creation and management of maxd databases
• input and export
• databases
• query-able by other software (e.g., StreptoBASE's
  maxdBrowse)
• allow MIAME-compliant data to be entered

27
maxd database structure

• Experiment
• Measurement
• Array Design
• Bio-Materials
• Protocols

28
maxd features

• INPUT
• forms
• bulk-upload from
• XL / tab-delimited files
• XML

• OUTPUT
• tab-delimited text
• MAGE-ML
• maxdml
• more in walkthrough...

29
StreptoBASE

• Your Streptomyces Gene Expression Resource a
  web front-end for maxd

30
Data warehouse portal

• Registration
• creates user account
• option to create user's maxd database
• Log-in via maxdLoad2
• data upload
• Log-in via browser
• browse data using maxdBrowse

31
maxdBrowse
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
Terms of Use

• Streptomyces microarrays
• from UniS Functional Genomics group
• can be obtained at a reduced price under
  Consortium arrangements
• Consortium committed
• to making results obtained using subsidised
  arrays publicly available

37
Terms of Use

• Data accessible
• immediately to Consortium-funded staff
• to help with upload, analysis and test software
• on publication to the world
• Analysis support
• for deposited data only
• merits either acknowledgement or authorship

38
Summary

• Registering on StreptoBASE allows
• to create a microarray database containing
• array design, source and protocols
• upload data via maxdLoad2
• MIAME-compliance
• browse data via StreptoBASE website
• more in walkthrough...

39
Future Pipeline
40
The Future

• For submitted data
• Data analysis pipeline
• Potential for complex queries
• Building these for you IMPOSSIBLE without your
  data, so PLEASE submit

41
Giles Velarde

• available for
• metadata upload training
• data analysis support and collaboration
• happy to visit or be visited
• likes kebabs and coffee
• is happy to help people build kebab recipe object
  models