1. Day 2: Session 8: Questions and follow-up…. James C. Fleet, PhD
Distinguished Professor
Department of Nutrition Science
Pete Pascuzzi, PhD
Assistant Professor
Purdue Libraries

2. Day 2 Session 9: Visualization I: Basic techniques James C. Fleet, PhD
Distinguished Professor
Department of Nutrition Science
Pete Pascuzzi, PhD
Assistant Professor
Purdue Libraries

3. What Data Do You Need? Data Needed Format Reason Raw Array Data
CEL file (Affymetrix)
Submit to GEO
Run analysis from beginning
Processed, normalized data
MAS5
RMA
(*.txt, Excel)
Present/Absent call
Primary data
Filtered to remove low expression genes
*.txt
Excel
Used for statistical analysis
Differential Expressed Gene List
*.gct
For GenePattern
For DAVID, Reactome, IPA
Fleet 2016

4. “learning a hidden data concept”
Clustering Methods
Grouping a large dataset into smaller data sets based upon some similarity
“learning a hidden data concept”
Marker selection-based:
Heat maps
Unsupervised learning/Class Discovery:
Connectivity-based: Hierarchical
Centroid-based: K-means
Self organizing maps
PCA
Karimpour-Fard et al. (2015) Hum Genomics Oct 28;9:28
Fleet 2016

5. Hierarchical Clustering Bottom-up or Agglomerative:
Start with all obs in one cluster then split recursively going down the hierarchy
Top-down or Divisive:
Bottom-up or Agglomerative:
Start with each obs = cluster then merge going up the hierarchy
Fleet 2016

6. Heat Maps https://www.youtube.com/watch?v=f5MhPVeA3Z4
Table of values where the value in a cell is color coded
Ordered so similar patterns are next to one another
Fleet 2016

7. K-Means Clustering (2) (1) (3) (4)
k clusters created by linking each observation to a mean
k initial means randomly generated
(2)
(1)
(3)
(4)
Repeat 2 and 3 until convergence
Centroid of each k clusters becomes the new mean
Fleet 2016

8. Self Organizing Maps (SOM)
Fleet 2016

9. Principle Components Analysis (PCA)
ID consecutive factors that are uncorrelated/orthogonal to each other
Scores Plot: used for interpreting relationships among observations
Scree Plot:
Defines the number of principle components worth considering
Account for less and less variability
Principle Component 2
Principle Component 1
Fleet 2016

10. a freely available open-source software package developed at the Broad Institute of MIT and Harvard for the analysis of genomic data.
Fleet 2016

11. GCT File Format for GenePattern
Fleet 2016

12. Tuesday
BREAK #1

13. Day 2 Session 10: Functional Assessment I:
Gene Set Enrichment Analysis
NIH David
James C. Fleet, PhD
Distinguished Professor
Department of Nutrition Science
Pete Pascuzzi, PhD
Assistant Professor
Purdue Libraries

14. …a collection of annotated gene sets for use with GSEA software
…a collection of annotated gene sets for use with GSEA software. 8 collections curated by the Broad Institute.
Subramanian et al. (2005) Proc Natl Acad Sci 102:15545.
Fleet 2016

15. Data format for Functional Classification
Huang et al. (2009) Nat Protocols 4:44.
Fleet 2016

16. Day 2 Session 11: Special Lecture Sean Davis, PhD Staff Scientist
National Cancer Institute

17. Day 2 Session 12: Visualization II: Intro to Pathway Analysis
> Reactome
> Pathview
James C. Fleet, PhD
Distinguished Professor
Department of Nutrition Science
Pete Pascuzzi, PhD
Assistant Professor
Purdue Libraries

18. Data Formats for Analysis
Fleet 2016

19. Data Format for Visualization
Data Format for Visualization
GeneID
DCIS_1
DCIS_2
DCIS_3
10000
10001
0.4159
10002
0.1985
0.0379
0.3419
10003
0.0364
10004
0.0018
10005
0.4778
10006
0.6503
0.1951
*.txt or *.csv format
Fold change or raw data
Also available in R as a Bioconductor package
Fleet 2016

20. Making Sense of Specific Gene Expression Data
BioGPS: Tissue Expression Pattern
Coremine: Gene Annotation
Fleet 2016

21. Tuesday
BREAK #2

22. Day 2 Session 13: Pathway Analysis continued…. James C. Fleet, PhD
Distinguished Professor
Department of Nutrition Science
Pete Pascuzzi, PhD
Assistant Professor
Purdue Libraries

23. Day 2 Session 14: Challenges of Metabolomic and Proteomic data
James C. Fleet, PhD
Distinguished Professor
Department of Nutrition Science
Pete Pascuzzi, PhD
Assistant Professor
Purdue Libraries

24. Levels of Metabolic Research
What may happen
Genomics ~25,000 genes
DNA (Genotype)
What can happen
Phenotype
Transcriptomics
~100,000 transcripts
What has happened
mRNA
Metabolomics
Lipidomics ~40,000
Metabolic pathway
Substrate
Product
Challenges:
Many metabolites
Location
Chemistry
Proteomics
~ a million proteins
What is happening
Adapted from Navas-Iglesias et al. (2009) Analytical Chemistry, 28, 393. by C. Ferreira (Purdue)

25. The Increasing Complexity of Omic Analysis
+ location, interaction, and chemistry differences
Genome
~25,000 genes
Transcriptome
~100,000 mRNA + miRNA + lncRNA
Proteome
~1,000,000
Alternative promoters, splicing, editing
Post-translationalmodification
Fleet 2016

26. Proteomics/Metabolomics Workflow
Data Filters
Intensity
Fold
Well Designed Study
Interpret Experiment
Normalized Vetted Data
Network building
Sample Preparation
Statistical Analysis
Sample Analysis and QC analysis
Pathway and Geneset Analysis
Link Peaks to Databases, Quantification, Normalization
Differentially Expressed List
Clustering and visualization
Raw Reads
Fleet 2016

27. Generalized Proteomic Pipeline
Steps 1-4 are distinct from the genomics pipeline
Walther and Mann (2010) J Cell Biol 190:491.
Fleet 2016

28. MaxQuant File formats MS-based peptide profile wifi RAW .d mzxml
File formats
wifi
RAW .d
mzxml
Link to peptide database
Integrate to proteins
Modified from Tiago Sobreira, Purdue
Fleet 2016

29. Proteomic Resources http://www.ebi.ac.uk/pride/archive/

31. Metabolomics Resources