InnateDB

1
InnateDB Facilitating Systems Level Analyses
of the Mammalian Innate Immune Response
David Lynn M.Sc., Ph.D., Research Associate,
Brinkman Lab., Simon Fraser University
Hancock Lab., University of British
Columbia. InnateDB Data Analysis Workshop -
UBC, Vancouver. April 2nd 3rd 2008. Updated
Sept. 2009.
2
Systems Biology Approaches to Investigating the
Innate Immune Response

• Although progress has been made in understanding
  the innate immune response including the detailed
  dissection of some of the critical signaling
  pathways involved.
• Now becoming clear that the innate immune
  response does not involve simple linear pathways
  but rather complex networks of pathways and
  interactions, negative feedback loops and
  multifaceted transcriptional responses.
• To better understand the complexities of the
  innate immune response and the cross-talk between
  its components, complimentary systems level
  analyses and more focused follow-up experimental
  approaches are now needed.

3
InnateDB Developed in the Context of Two Large
International Systems Biology Projects
Mouse Model Datasets Cerebral Malaria mouse
model (IMR, Australia) Tuberculosis mouse model
(AECM) Shigella xenograft model (Pasteur) Human
Clinical Datasets Typhoid Malaria Vietnam
(OUCRU/Stanford/Sanger) Non Typhoidal Salmonella
Malawi (Sanger) Chronic/Acute Helminth Ecuador
(USF de Quito/Sanger)

Modulating innate immune response viaHost
Defense peptides (Hancock lab, UBC) Mouse KOs
(Sanger)
Novel insight into host response and mechanism of
peptides. Common Pathways, networks and
transcriptional regulation?
4
Why Systems Approaches are Needed

• Many layers of complexity
• Layers of regulation
• 100s 1000s DE genes
• Not simple pathways ? networks of molecular
  interactions.
• Gardy, Lynn, Brinkman, Hancock.
• Enabling a systems biology approach to
  immunology focus on innate immunity.
• Trends in Immunology June 2009.

5
The Need for InnateDB the Manual Curation of
Innate Immunity Relevant Molecular Interactions
Pathways.

• Quickly apparent that available resources
  provided poor coverage and detail of the
  molecular interactions and pathways relevant to
  innate immunity.
• This information is essential for the
  systems-orientated interpretation of large scale
  genomics data.
• TLR4 ? one of the most important molecules in the
  innate immune response, has relatively few
  molecular interactions annotated in the major
  publicly available interaction DBs.
• 5 of these DBs combined contained annotated
  molecular interactions between TLR4 and just 11
  other proteins.
• Through a review of the literature we have
  curated, in detail, a further 16 unique
  interactions, and provided annotation of nearly
  60 different lines of evidence supporting these
  interactions.
• Relatively new pathways (NLR, RLR pathways) not
  annotated at all in major pathway databases.
• Few resources available for analysis of data in a
  pathway/network context that were accessible to a
  biologist. No resources for innate immunity.

6
Overview of InnateDB Project (www.innatedb.ca)

• InnateDB (www.innatedb.ca) is a database of all
  human and mouse experimentally-verified
  interactions and pathways
• ( their component molecules
  Genes/Proteins/RNAs).
• Particular emphasis on the contextual manual
  curation of interactions involved in innate
  immunity (10,000 intxns).
• InnateDB facilitates systems-level analyses of
  mammalian signaling through integrated
  bioinformatics and visualization tools pathway
  ontology analysis, network construction
  analysis, orthologs, Cerebral, Cytoscape, CyOOg,
  etc.
• Manual curation project integration of publicly
  available databases into InnateDB greatly
  increases innate immunity relevant molecular
  interaction networks pathways.
• Enable biologists without a computational
  background to explore their data in a more
  systems-oriented, yet user-friendly, manner.

7
Contextually Curating Innate Immunity-Relevant
Interactions

• Manual curation gt 10,000 innate immune-relevant
  interactions (human and mouse).
• Involving 2,700 genes from review of 2,600
  unique publications.
• We can often double of interactions for a given
  gene.
• Pathways interactions are curated with
  contextual annotations
• (supporting publication participant molecules
  the species the interaction detection method
  the host system the interaction type the cell,
  cell-line and tissue types etc).
• Developed InnateDB submission system software to
  allow submission of interaction annotation in an
  ontology-controlled and MIMIx PSI-MI 2.5
  compliant manner.
• Developed curator tool software to allow curators
  modify existing annotations.

8
Going Beyond Innate Immunity A Centralized
Resource for Interactions Pathways

• Aside from the well known signalling pathways ? a
  range of other disparate processes, including
  apoptosis, ubiquitination, endocytosis, cell
  activation and recruitment ? all required to
  mount effective innate immune response.
• Adding to this complexity ? borders between the
  innate and adaptive immune responses are becoming
  increasingly blurred.
• Furthermore, if we hope to identify new networks
  or pathways involved in innate immunity, analyzes
  must include genes and proteins that are, as yet,
  not known to play specific roles in the innate
  immune response.
• To address these issues ? InnateDB also
  incorporates data on the entire human and mouse
  interactomes.

9
Going Beyond Innate Immunity An Integrative
Biology Resource

• 115,000 human and mouse interactions extracted
  loaded from BIND, INTACT, DIP, BIOGRID MINT
  DBs.
• Cross-referenced genes to gt3,000 pathways from
  KEGG, PID, BIOCARTA, INOH, NetPath Reactome
  DBs.
• Allows one to visualize/analyze interactions
  associated with specific pathway.
• Pathway ORA.
• Annotation from Ensembl provides details of human
  mouse genes, transcripts and proteins.
• UniProt, Entrez, Gene Ontology ? rich protein
  gene annotation.

10
Through manual curation integration of existing
data from publicly available databases we can
greatly increase innate immunity relevant networks
TLR4 direct and secondary interactions annotated
by InnateDB
TLR4 direct and secondary interactions annotated
by MINT Database
11
Direct and Secondary Interactions of TLR4 in
InnateDB(20 of these interactions unique to
InnateDB)
12
www.innatedb.ca
13
InnateDB Advanced Yet User-Friendly Searching
Find Analyze Relevant Interactions, Pathways
Genes/Proteins.
14
InnateDB Facilitating Systems-Level Analyses of
Gene Expression Data
Upload Your Own Gene Expression Data - Up to 10
conditions/timepoints at 1 time.
Overlay Gene Expression Data from Multiple
Conditions on Networks/Pathways
Pathway, Gene Ontology TF ORA tools Find DE
Pathways/Functionally Related Genes/TFs
Go Beyond Pathway Analysis Differentially
Expressed Sub-networks New Pathways? How Are DE
Genes Actually Inter-connected? Central
Regulators (Network Hubs)
15
Pathway Analysis Any type of Quantitative Data.
Orthologous Pathways
GWA Candidate Associated Genes

• InnateDB pathway analysis
• identify OR pathways.
• highlight potentially unknown relationships
  between makers on different chromosomes.

16
Constructing Analyzing Networks Using InnateDB

• Pathway analysis can be very powerful in
  determining which annotated pathways are most
  significantly associated with DE genes.
• Network analysis ? move from simple view of the
  signaling response to a more comprehensive
  analysis of the molecular interactions between DE
  genes and their encoded proteins RNAs.
• Potentially uncover as yet unknown signaling
  cascades or pathways, functionally relevant
  sub-networks and the central molecules, or hubs,
  of these networks.

17
Results Visualize Gene Expression Data in an
Interaction Network Context
18
Multi-experiment View in Cerebral
19
Robust Orthology Gene Order Predictions
Facilitating Comparative Analysis

• Majority of mammalian interaction data available
  in InnateDB and other interaction databases
  primarily refers to human genes and proteins.
• To facilitate comparative network-based analysis
  of the human, mouse and bovine interactomes,
  detailed orthology predictions have been
  integrated into InnateDB.
• Orthology predictions generated using an in-house
  method, Ortholuge, which provides accurate
  predictions of orthology using a phylogenetic
  distance-based approach.
• Orthology predictions are further supported
  through the development of a human and mouse gene
  order and synteny browser.

20
A Guide to Using InnateDB
21
InnateDB User Friendly Interface www.innatedb.ca
22
(No Transcript)
23
Not sure what you want to search for? Browse
InnateDB by Interaction Type, Pathway or Various
Immune Gene Lists
24
All InnateDB Interactions Can be Downloaded in
Proteomics Standards Initiative (PSI) 2.5 XML
Format
25
Resources Page Details of Relevant Software,
Databases, and Immune gene Lists
26
Statistics on Curated Interactions Interactions
from other Databases
27
Use contact form or send email to
innatedb-mail_at_sfu.ca to report bugs, errors or to
get involved in curation.
28
Documentation, Tutorials Help
29
Searching InnateDB
30
Do a simple search for genes, proteins or
interactions of interest on the InnateDB hompage
e.g. IRAK genes.
31
Advanced Search for Genes Proteins
32
Advanced Search for Interactions
InnateDB contains detailed information for more
than 115,000 human and mouse molecular
interactions integrated from several of the major
public interaction databases along with 10,000
manually-curated innate immunity relevant
interactions.
To reduce redundancy, interactions in InnateDB
that have the same participants and interaction
type are grouped together by default. Choose 'No'
to return all redundant interactions separately.
33
Search for Particular Interactions or Genes that
are in a Specific Pathway
34
Search Results searching for genes of interest
e.g. IRAK
35
Search Results searching for genes of interest
e.g. IRAK
36
Interaction Results Page.
37
(No Transcript)
38
Visualize Interactions in a subcellular
localization-based layout using the Cerebral
plugin for Cytoscape.
39
How a biologist thinks of a pathway .
40
Pathway Visualization in Cytoscape
41
Pathway Visualization using Cerebral
www.pathogenomics.ca/cerebral (Bioinformatics
2007)
42
A Quick Guide to Using Cerebral in InnateDB

• Cerebral can be used to visualize interaction
  networks from a set of interactions from
  InnateDB.
• Cerebral uses subcellular localization
  annotations to provide more biologically
  intuitive pathway-like lay-outs of interaction
  networks.
• Note the subcellular localizations in Cerebral
  should only be used as a guide. There are many
  proteins with no annotated subcellular
  localizations and many others that have multiple
  possible localizations (only 1 will be shown,
  nuclear, extracellular and membrane localizations
  will take precedence over cytoplasm if there are
  multiple).
• InnateDB batch searching allows users to upload a
  list of genes along with associated gene
  expression data from up to 4 different
  conditions.
• Gene expression data can be overlaid on network
  data and you can visualize this in Cerebral.

43
Opening Interaction Data in Cerebral from an
Interaction Results page in InnateDB.

• You will be prompted to open a .jnlp file.
• You are recommended to save this file to your
  computer and then open it this will allow you
  save a copy of this dataset.
• Opening the .jnlp file directly without saving
  sometimes causes Cerebral to hang when loading
  large datasets.
• Note to use Cerebral you need to install Java
  version 6 or greater.
• You can get this from http//java.com/en/download/
  index.jsp

44
Opening Cerebral

• Cerebral is a Java plugin for the Cytoscape
  Visualization software.
• When you open the .jnlp file Cytoscape will begin
  downloading.
• You will then be prompted Do you want to run
  the application click Run.

45
Cerebral is Now Open and Displays Interactions
Based on Protein Subcellular Localizations
46
Re-size the Network
Click here to re-size the network display to
full-screen.
47
Navigating in Cerebral

• Right click and push your mouse forward or back
  to zoom.
• Hold middle button of your mouse and drag to
  navigate around the network.
• Grey nodes do not have an annotated subcellular
  localization (from Gene Ontology data in
  InnateDB).
• Lines connecting nodes represent interactions.
  Dashed lines have only 1 supporting publication
  in InnateDB. The thicker the line the more
  publications support the interaction.

48
Interactively Link back to InnateDB to Look up
Information on Particular Genes/Interactions of
Interest.

• Right-click on a node (protein/gene) or edge
  (interaction line) to link to the relevant gene
  or interaction details page in InnateDB.

49
Nodes Can be Dragged to Other Layers as Desired.
50
Do a simple search for genes, proteins or
interactions of interest on the InnateDB hompage
e.g. IRAK genes.
51
View Detailed Gene Annotation.
52
Gene Details Page.
53
Gene Details Page Molecular Interactions Gene
Ontology Annotation.
54
Integrated Orthology Gene Order Information
55
Human/Mouse Conserved Gene Order Synteny Browser
56
Gene Details Page Associated Pathways
57
Gene Details Page Cross-references to other
Databases
58
Integrating Gene Expression Data in a Molecular
Interaction Network and Pathway Context
59
InnateDB Integrating Gene Expression Data in a
Molecular Interaction Network and Pathway Context
Integrated Gene Expression Data with Molecular
interaction data Pathway associations Rich gene
annotation
Batch Search of InnateDB
Microarray Data ? Differentially expressed Genes
60
Orthologous Interaction Networks

• Detailed protein/gene interaction data mainly
  available for human.
• Can use InnateDB ortholog predictions in mouse
  and cow
• Build the hypothetical orthologous interaction
  network for genes of interest in these species.
• Find associations to pathways for orthologous
  genes e.g. map pathways to mouse genes based on
  human orthology.
• Predict potential differences in different
  species e.g. missing orthologous gene in one
  species ? may indicate reliability as model
  organism for network of interest.
• Compare orthologous predicted networks to
  experimental data e.g. in mouse.

61
Example Tab-delimited File
62
Upload Gene/Protein List to InnateDB Along with
Any Associated Quantitative Data
Select a file to upload by clicking on the
"Upload File" button - upload a tab-delimited
file of protein/gene identifiers or accession
numbers and obtain a list of all genes, proteins,
pathways, interactors or interactions that they
are associated with. Alternatively, click on the
"Web Form" button and paste your tab-delimited
data in the text box (max. 1000 lines)
63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
Results Visualize Gene Expression Data in an
Interaction Network Context
68
Multi-experiment View in Cerebral
Click on one of the mini-windows to view data for
condition in large window.
69
Cerebral
Multi-Array Viewer
70
Cerebral Multi-Array Viewer
71
Interactively Link back to InnateDB to Look up
Information on Particular Genes/Interactions of
Interest.
72
Pathway Over-representation Analysis
73
Return Pathways Associated with Uploaded Gene
List

• To do pathway over-representation analysis (ORA)
  you first need to upload a list of gene
  identifiers and associated fold-change in gene
  expression values (and P values) as described
  above.
• InnateDB recommends that you to upload All genes
  from your array dataset not just differentially
  expressed (DE) genes (probes mapping to multiple
  different genes should be removed). The pathway
  ORA tool uses the proportion of DE genes on the
  whole array to determine if a particular pathway
  is significant.
• As the above method can be very conservative due
  to the large number of tests performed -InnateDB
  also provides users with the option of uploading
  a subset of genes and performing the pathway ORA
  analysis. This subset analysis uses a slightly
  different algorithm that does not take gene
  expression values into account. This is necessary
  as the algorithm does not know the proportion of
  DE genes on the array. Therefore, this analysis
  cannot handle data from multiple conditions.
• If you have multiple probes for the same gene
  these values will be averaged for the purposes of
  the pathway ORA.
• Because InnateDB sources its pathway data from
  multiple databases, each with its own
  interpretations of the components of a given
  pathway, you will observe some degree of
  duplication in the results however, this is
  outweighed by the extra annotation that can be
  obtained from different data sources.

74
Pathways Associated with Uploaded List
75
Choose Parameters for Pathway ORA
Choose fold-change in gene expression threshold
(determines which genes are considered
differently expressed) Default /- 1.5.
Choose P value threshold associated with each
fold-change in gene expression value. (determines
which genes are considered differently expressed)
Default P lt 0.05. Several different
statistical methods are available to determine if
pathways are significantly associated with DE
genes - Hypergeometric, Fisher Chi Square.
Two options to correct for multiple testing are
included - The Benjamini Hochberg correction
for the FDR and the more conservative Bonferroni
correction.
76
(No Transcript)
77
Pathway Summary Page
KEGG pathway diagrams can be dynamically linked
to overlaying gene expression data
78
Acknowledgements The Bioinformatics Team

• Overall Project Management
• Bob Hancock
• Brett Finlay
• Lorne Babiuk
• Bernadette Mah
• Bioinformatics InnateDB Management
• Fiona Brinkman
• David Lynn
• InnateDB Database Development/Data Loading
• Matthew Laird
• Nicolas Richard
• Fiona Roche
• Timothy Chan
• Michael Acab
• InnateDB Search Engine User Interface
• Geoff Winsor

• InnateDB Submission System Curator tool
• Calvin Chan
• Naisha Shah
• Cerebral Pathway Visualization Software
• Jennifer Gardy
• Aaron Barsky
• Tamara Munzner
• Orthologs Gene Order
• Dan Tulpan
• Matthew Whiteside
• Mark Sun
• Matthew Laird
• Matthew Whiteside
• Systems Administration
• Matthew Laird (SFU)
• Timothy Chan (UBC)