1. The Computational Biology and Informatics Laboratory
An Informatics Framework for the Analysis of Gene Regulation and Pathways
The Computational Biology and Informatics Laboratory

2. Gene Regulation Stem Cell Development Organismal Biology
Tissue specificity
Developmental regulation
Stem Cell Development
Pancreatic islet cells
(hematopoiesis/ erythropoiesis/ others)
Organismal Biology
Plasmodium falciparum
mouse chromosome 5/neural crest

3. Examples of Systems Under Study
Endoderm
Pancreatic anlage
PDX-1
p48
Pax4
Exocrine
Endocrine

4. Knowledge Domains Sequence/ Sequence annotation Pathways/
Gene expression
experiment
Proteomics,
Metabolomics
Pathways/
Networks

5. Knowledge Domains Sequence/ Sequence annotation Pathways/
Gene expression
experiment
Proteomics,
Metabolomics
Pathways/
Networks

6. GUS: Genomics Unified Schema
free
text
Controlled vocabs. GO
Species
Tissue
Dev. Stage
Genes, gene models
STSs, repeats, etc
Cross-species analysis
Genomic
Sequence
RAD
RNA Abundance DB
Characterize transcripts
RH mapping
Library analysis
Cross-species analysis
DOTS
Transcribed
Sequence
Special
Features
Transcript
Expression
Arrays
SAGE
Conditions
Ownership
Protection
Algorithm
Evidence
Similarity
Versioning
under development
Domains
Function
Structure
Cross-species analysis
Protein
Sequence
Pathways
Networks
Representation
Reconstruction

7. What is GUS ? A relational schema A Perl API and annotation subsystem
Over 180 tables
Organized around central dogma
A Perl API and annotation subsystem
lightweight object layer with “plug-ins”
supports high-level programmatic access but allows SQL
A generic user interface
Java Servlet-based (Apache JServ)
supports browsing and also restricted ad-hoc queries
A data warehouse
GenBank, dbEST, SWISS-PROT, UCSC “Golden Path”, others
Gene Ontology (GO) terms and assignments
Controlled vocabularies: taxonomy, anatomy, disease state
DoTS: database of assembled ESTs and mRNAs

8. Clusters vs. Contig Assemblies
UniGene
Transcribed Sequences (DOTS)
BLAST: Clusters of
ESTs & mRNAs
CAP4: Consensus Sequences
-Alternative splicing
-Paralogs

9. Mouse Assemblies Over 2 million mouse EST and mRNA sequences used
(loaded into GUS as of June 1, 2001)
Combined into 367,525 assemblies
71,602 assemblies had more than one sequence
22 sequences on ave./non-singleton assembly
993 nt = ave. length of non-singleton assemblies

10. Assembly Validation Alignment to Genomic Sequence via Blast/sim4.
preliminary data look good
Assembly consistency (Assemblies provide potential SNPs)
Add BLAST sim4 figure

11. Predicting Gene Ontology Functions

12. GUS Annotation Interface

13. Knowledge Domains Sequence/ Sequence annotation Pathways/
Gene expression
experiment
Proteomics,
Metabolomics
Pathways/
Networks

14. RAD Multiple labs Multiple biological systems Multiple platforms
Expressed genes?
Differentially-expressed genes?
Co-regulated genes?
Gene pathways?

15. RAD: RNA Abundance Database
Experiment
Platform
Raw Data
Processed
Data
Algorithm
Metadata
Compliant with the MGED standards

16. Microarray Gene Expression Database group (MGED)
International effort on microarray data standards:
Develop standards for storing and communicating microarray-based gene expression data
defining the minimal information required to ensure reproducibility and verifiability of results and to facilitate data exchange (MIAME, MAGEML-MAGEDOM)
collecting (and where needed creating) controlled vocabularies/ ontologies.
developing standards for data comparison and normalization.
The schema is compliant with the minimum annotations recommended by MGED.
MIAME: Minimum Information About a Microarray Experiment (common set of concepts that need to be captured in a database to describe gene expression experiments adequately for interpretation, reproduction or critical assessment).
MAML: MicroArray Mark-up Language (XML Document Type Definitions of the concepts).

17. Experiment Tables Label Sample Treatment Disease Devel. Stage Anatomy
Hybridization
Conditions
Label
Sample
Treatment
Disease
Devel. Stage
ExperimentSample
Anatomy
Taxon
RelExperiments
Exp.ControlGenes
ControlGenes
Experiment
ExpGroups
Groups

18. Query RAD by Sample or by Experiment
Access by Experiment
groups
Sample info
ontologies
Image info

21. Storing the Quantified Data is Just the Beginning
Analysis result
e.g., cluster #
is differentially-
expressed
Output of
image analysis
software
Normalized
data
Selected data
for analysis
SpotResult/
SpotFamilyResult
tables
Analysis/ Algorithm tables

22. Different Views of GUS/RAD
Focused annotation of specific organisms and biological systems:
organisms
biological systems
Endocrine
pancreas
Human
Mouse
CNS
GUS
GUS
Plasmodium
falciparum
Hematopoiesis
*not drawn to scale*

23.
New site

24. Contig View OM Restriction Sites Microsatellites Self-BLAST NRDB-BLAST
SAGE Tags
EST/GSS
FullPHAT
GeneFinder
GlimerM
Annotation (chr2-TIGR)

25. Gene Page - I Description Notes Protein Graphical View
Genomic Neighborhood GV
P. yoelii similarity
NRDB
ProDom

26. Protein Graphical View
BLASTP
Secondary Structure
Xmembrane
Motifs
Signal Peptides
Hydropathy

27. Boolean Queries

28. AllGenes

29. Assembly/RNA View

30. The Gabrg1-Gabra2-Gabrb1-Txk-Tec-Gsh2-Pdgfra-Kit-Kdr(Flk1)-Clock BAC contigs on Chr. 5
Sequence available
Sequence available

31. DoTS Assemblies Can Provide A Bridge Between Radiation Hybrid and BAC Contig Maps
RH Map
AV026557
AI848177
AI132477
AW490897
AI507113
AV038945
AV074028
C85052
AV364670
AW987574
AF026073
AF022894
AI586015
C80280
Kit

32. Annotation of Mouse BAC Draft Sequence: Localization of the mouse corin gene
Update?

33. Annotation of Kit draft sequence (232h18) Ordering and orienting pieces using conserved regions

34. Annotation of Kit draft sequence (232h18) Transcription Element Search System analysis
Searched entire human and mouse orthologous sequences with all TESS matrices.
Identified binding sites over/under-represented in the conserved regions.
Conserved sites dispersed over 150kb.
Over-represented factors include AP2, Pax-6, S8, Oct-1, E2A, E2F-DRTF, TAL1-/E47, CdxA, Ubx, AbdB-r, Engrailed, Hairy, DFD

38. Connecting Genes and Gene Expression to Pathways

40. The allgenes (GUS) index provides annotation of array elements in RAD
EST clustering
and assembly
Different representations of the same RNA are identified.
EST/mRNA annotations are combined.
Consensus sequence is annotated (e.g., gene function).

41. Creating a “pancreas chip”
Top 15% of clone signals in 2 mouse pancreas, 1 human islet, and 1 human insulinoma (GEM) array experiments.
AND
All the ESTs from 5 islet cDNA libraries.
Find mouse and human RNAs in GUS containing these clones/ESTs.
If human, BLAST against mouse RNAs to find ortholog.
Non-redundant list of mouse RNAs
List of mouse IMAGE clone IDs

42. RAD GUS EST clustering and assembly Identify shared TF binding sites
TESS (Transcription Element Search Software)
Genomic alignment
and comparative
Sequence analysis
Identify shared
TF binding sites

43. Example of Systems Under Study: Pancreatic development
Endocrine
progenitor
Exocrine
exocrine cell
PDX-1
PDX-1 Ngn3
Beta2 (NeuroD)
Common progenitor
PDX-1 Pax4
Nkx6.1
Beta2 Isl1
Pax6 Nkx2.2
Brn 4
alpha cell
PP cell
Glucagon
beta cell
delta cell
Pancreatic
polypeptide
Insulin
Somatostatin
Beta/Delta
Alpha/PP
p48/PTF1
Amylase
Adapted from: Huang Tsai, J Biomed Sci 2000:7:27-34 and Jensen et al, Diabetes 2000:

44. CAP4 provided by Paracel
Acknowledgements
CBIL: Chris Overton
Chris Stoeckert
Vladimir Babenko
Brian Brunk
Jonathan Crabtree
Sharon Diskin
Greg Grant
Yuri Kondrakhin
Georgi Kostov
Phil Le
Elisabetta Manduchi
Joan Mazzarelli
Shannon McWeeney
Debbie Pinney
Angel Pizarro
Jonathan Schug
PlasmoDB collaborators:
David Roos Martin Fraunholz
Jesse Kissinger Jules Milgram
Ross Koppel, Monash U.
Malarial Genome Sequencing Consortium
(Sanger Centre, Stanford U., TIGR/NMRC)
Allgenes.org collaborators:
Ed Uberbacher, ORNL
Doug Hyatt, ORNL
EPConDB collaborators:
Klaus Kaestner Marie Scearce
Doug Melton, Harvard
Alan Permutt, Wash. U
Comparative Sequence Analysis Collaborators:
Maja Bucan Shaying Zhao
Whitehead/MIT Center for Genome Research
CAP4 provided by Paracel

45. WWW.CBIL.UPENN.EDU “allgenes” human and mouse gene index:
PlasmodiumDB:
RAD, RNA Abundance Database:
Endocrine Pancreas Consortium Database:
TESS, Transcription Element Search System
PaGE, Patterns from Gene Expression
MGED:

46. Summary
Genomics Unified Schema (GUS) integrates and adds value to genomic, transcribed, and protein sequence.
RNA Abundance Database (RAD) captures experiment, platform, data, and analysis from array and SAGE experiments. RAD adds value through integration with GUS.
System-specific views are available for human and mouse, Plasmodium falciparum, and endocrine pancreas.
GUS and RAD can be used to design custom arrays, identify potential SNPs, genome annotation, and comparative sequence analysis.
Tools such as TESS and PaGE have been developed to analyze the data in GUS and RAD.
Include TESS? Add simulations result. Applying to microarray - issue of cleaning up and normalization

47. EST libraries (wide coverage, low resolution)
Ontologies
(anatomy, development,
disease)
Expression patterns
Microarrays
(high resolution)
Expression rules
TFBS (promoter analysis)
Protein domains (splice forms)

48. Ontologies in Gene Expression Databases
Controlled vocabulary (ontologies not always needed)
hierarchical
Directed acyclic graphs
Schema
Concepts as objects or relational tables
Attributes and data types provide specification
Relationships specified through subclassing (objects) or foreign keys (relational tables)
Knowledge representation
Link to other domains (gene sequence annotation, gene and protein roles, pathways)
Facilitate data exchange by mapping common concepts

49. GUS Object View Gene Gene Feature Genomic Sequence NA Sequence RNA RNA
Protein
Protein
Feature
Protein
Sequence AA
Sequence

50. High Level Flow Diagram of GUS Annotation
Genomic
Sequence
mRNA/EST
Sequence
BLAST/SIM4
ORNL Gene predictions
GRAIL/GenScan
Clustering and
Assembly
Predicted
Genes
DOTS consensus
Sequences
Merge Genes
Gene/RNA cluster
assignment
Gene
Index
Gene families, Orthologs
Assign Gene Name,
Manual Annotation..
Predicted RNAs
Predicted Proteins
Grail/Genscan,
DIANA/framefinder
BLASTX
PFAM,SignalP,
TMPred, ProDom, etc
BLASTP
Algorithms for
functional predictions
BLAST Similarities
Protein
Features/Motifs
GO Functions
CellRoles

51. Anatomy Hierarchy

52. Predicted GO Functions

53. Schema Browser

54. Summary of allgenes.org content
Update, add GO figure

55. Information to be captured
Figure from:
David J. Duggan et al. (1999) Expression Profiling using cDNA microarrays. Nature Genetics 21: 10-14

56. The Purpose of GUS
Integrate >> Annotate >> Mine (and Track)
Integrate existing databases and tools
a single point of access to what is already known
Provide an automated “lab. notebook”
a permanent record of work in progress
e.g., similarity searches, array data, etc.
And ultimately: support data mining
a potential source of novel discoveries

57. Query “History” Feature1 2 4 3

58. Critical Assessment of Microarray Data Analysis ‘00
Golub et al. (1999), Science, 286:
ALL-AML: heterogeneous groups:source (B-cells, T-cells, 4 AML types), sex, success, etc.
Focus on B-cells (37 replicates) vs T-cells (9 replicates): combined the training and the test sets
Affymetrix
single sample hybridization
each signal is a composite of hybridizations to probes in a set
absent calls
Note that the data was already normalized.

59. Distribution Heterogeneity
From the B-cells in the Golub et al. dataset.

60. “Deterministic” differential expression
B and T B T
Deterministic picture: this one shows a (nearly) deterministically differentially expressed gene. If not for the one absent call in the T-cell distribution, it would be deterministic.
log scale
Identifier: U23852, T-lymphocyte specific protein tyrosine kinase p56lck (lck) aberrant mRNA

61. “Non-deterministic” differential expression
B and T B T
log scale
Identifier: M23323, T-cell surface glycoprotein CD3 epsilon chain precursor

62. PaGE: Patterns from Gene Expression
PaGE assigns confidence measures to predictions of differential expression. Handles multiple testing in a nonparametric (and non-standard) way.
Does not use t-statistic.
Patterns are generated by comparison of groups of replicates to a reference group.
See Manduchi et al., Bioinformatics 16: , 2000.

63. PaGE: outline Find C (the upper cutratio) such that
is small (this is the false positive rate). Here i varies too.
This C gives a cutoff for making predictions about up-regulation.
Similarly for down-regulation (find an appropriate c [lower cutratio], reverse the above inequalities).

64. PaGE: approximations is bounded above by
After having shifted all intensities by an appropriate numerical constant, we approximate the unknown distribution of by that of
where i varies over the gene tags and j varies of the replicates for group 1. Similarly for group 2.

65. B-cell vs. T-cell using PaGE
Here a confidence of 0.90 is used, the false positive rate is around So that on the shift error tables, you can see that a shift of 5000 and a f.p.r. of is about optimal across all tables.
Column n,9 = fraction of times gene is up-regulated in T-cells out of 100 comparisons between n randomly chosen B-cell and all 9 T-cell expmts.

66. B-cell vs. T-cell using t-statistic
Column n,9 = fraction of times gene is up-regulated in T-cells out of 100 comparisons between n randomly chosen B-cell and all 9 T-cell expmts.