Bioinformatics%20in%20the%2090

1
Bioinformatics in the 90s

• Origins data storage needs related to the
  sequencing effort...
• but storage was hardly enough additional
  needs
• Assembly, comparison and annotation of sequences
• Prediction of genes
• Reconstruction of evolutionary trees
• Modelisation prediction of 3D structures
• ...
• IT on-line databases and software tools
• Science modeling, computational
  representations, algorithms

2
The post-genomic phase transition

• Availability of complete genome sequences
• High-throughput experimental techniques yield new
  types of results
• SNP (Single Nucleotide Polymorphisms)
• mRNA expression levels (DNA chips)
• Systematic determination of 3D structure
• Protein expression levels
• Protein- protein interactions
• Systematic mutagenesis
• ...
• New needs opportunities
• Processing and analysis of each type of data
• Integration of heterogeneous data
• Reconstruction and simulation of cellular
  mechanisms

3
Corporate Information
Founded December 1997 Headquarters and
laboratories Central Paris Employees 60
as of end 2001 Intellectual Property 57
patents on technology, interaction
s and targets Equity raised c. ?30
million Ownership Advent (B), Alafi
(US), Apax (F), Auriga (F),
IMH(D), Health Cap (S), Lombard-Odie
r (CH), Medicis (D), Rendex (B)
4
Hybrigenics business strategy

• Own drug discovery programs
• in the fields of infectious diseases, cancer and
  metabolic disorders
• the resulting novel validated targets being
  exploited for the Companys own product pipeline
• Collaboration and licensing agreements with
  biopharmaceutical companies
• in any disease field
• for out-licensing

5
Hybrigenics discovery programs
Cancer Proteins involved in basic cellular
functions Proteins involved in
apoptosis Proteins involved in cell cycle
regulation
Metabolic disorders / Obesity Proteins
involved in adipogenesis
Anti-infectious diseases
Antibacterial Essential proteins of the
pathogens HIV, HCV protein-protein interactions
between the host cell and the pathogen
6
The Helicobacter pylori Genome
From Tomb et al. (1997), Nature 388539-47
less than 20 with assigned biological
functions (500 with no database match 250 with
structural homology but totally unknown function)
1,667,867 base pairs 1,590 predicted ORFs
7
The protein-protein interaction map of
Helicobacter pylori
285 baits 261 proteins
2 million prey fragments
20 milion interactions/bait
PBS filtering (false positives identification)
Over 1,200 interactions Over 1500 SID
Nature (2001) 409211-215.
Connectivity 46.6 of proteome 3.36
interactions/bait Reproducibility gt95
8
Target IdentificationHybrigenics' PIM Technology
Platform
New Generation of Reliable High-Throughput 2-Hybri
d in Yeast Coli
PIMBuilder in-house Production Management System
PBS Scoring Technology
VirtualPIM Prediction
PIMRider platform
9
HybrigenicsTarget Discovery Process
Target Identification
Target Pre-Validation
Target Validation
Selected Pathology and Mechanism of Action
10
In-silico Target Validation Platform

• Goals
• Validate protein interactions and SIDs
• Evaluate  target potential  and druggability
• Provide functional context for target candidates
• Prioritize  promising" candidates for biological
  validation

• Means
• Integrate PIMs with functional clues of different
  origins
• Predict novel biological information
• Computer aided decision process
• Provide comprehensive  decision-oriented  view
  of functional clues
• Automated filtering

• Output
• Prevalidated targets functional context

11
The Genostar platformA modular software platform
for exploratory genomics
The Geno Consortium Pasteur Institute
(Paris), National Institute for Research in
Computer Science (INRIA, Grenoble) Genome Express
(Grenoble) Hybrigenics

• Genostar technology
• Rich object-based knowledge representation system
  (objects, relations, tasks and strategies)
• Modular architecture
• Domain-specific biological modeling

12
Genolink viewing biological data as a graph of
relations
Genolink Composite Graph
Vertices biological entities
Edges similarity, interaction or association
links
Sequence Similarity Links
Profile Similarity Links
Domain Inclusion Links
Tissue Expression Links
Subcell Location Links
Protein Interaction Links
Preprocessing
Genomic data
mRNA Expression data
Interaction data
Sub-Cellular Location
Domain data
13
From PIMs to Pathways
From PIMs to Pathways
Combine PIMs and external data to reconstruct
biological pathways
PIM annotation Pathways expansion
PIM Network of interaction links
Context-dependentHomology
Common Data Model
Functional Classifications
Pathways Databases
PIMs
14
The BioPathways Consortium

• Mission
• Foster development of pathways informatics
  systems biology
• Goals
• Scientific community buildup, standards
  recommendation, public outreach,
  industry-academia collaboration support,
  coordination with other groups
• Means
• Forum open to interested participants (academics,
  pharmas, biotechs, software vendors)
• Achievements
• Launched June 2000 by 3rd Millennium (Boston) and
  Hybrigenics (Paris)
• 1st Meeting at ISMB 2000 -gt Work Groups
• 2nd Meeting at PSB 2001 -gt First results on
  evaluation of pathways representations
• 3rd Meeting Satellite Meeting of ISMB 2001,
  Copenhagen -gt Focus on ontologies and pathways
  reconstruction (gt150 attendants), new workgroups
• Several sponsors (pharmas, biotechs, IT
  companies)
• Over 200 participants from academia industry

15
Annotation fonctionnelle

• Objectif assigner une/des fonction(s) à un
  gène ou à une protéine de séquence connue
• Méthodes traditionnelles
• Résultats expérimentaux
• Variations sur le thème propagation
  dannotations dorigine expérimentale via
  similitude de séquences
• Fonction ?
• Locale et précise (Ex la protéine P est un
  enzyme catalysant la réaction R)
• Globale et vague appartenance à un processus
  biologique de haut niveau (Ex P intervient dans
  la dégradation du glucose)
• Ce qui est propagé mots clefs, nœud dun arbre
  de classification fonctionnelle

16
An effort toward consensus Gene Ontology
Fig. 1 Examples of Gene Ontology. Three
examples illustrate the structure and style used
by GO to represent the gene ontologies and to
associate genes with nodes within an ontology.
The ontologies are built from a structured,
controlled vocabulary. The illustrations are the
products of work in progress and are subject to
change when new evidence becomes available. For
simplicity, not all known gene annotations
have been included in the figures. a, Biological
process ontology. This section illustrates a
portion of the biological process ontology
describing DNA metabolism. Note that a node may
have more than one parent for example, DNA
ligation has three parents, DNAdependent DNA
replication, DNA repair and DNA recombination
. b, Molecular function ontology. The ontology is
not intended to represent a reaction pathway, but
instead reflects conceptual categories
of gene-product function. A gene product can be
associated with more than one node within an
ontology, as illustrated by the MCM proteins.
These proteins have been shown to bind chromatin
and to possess ATPdependent DNA helicase
activity, and are annotated to both nodes. c,
Cellular component ontology. The ontologies are
designed for a generic eukaryotic cell, and are
flexible enough to represent the known
differences between diverse organisms.
The Gene Ontology Consortium (2000) Nature Genet.
25 25-29
17
Le dogme
Séquence
Structure
Fonction
18
et les expériences
Contexte cellulaire
Technologies de Perturbation
?
Séquence
?
Structure
Technologies dobservation
?
Fonction
Phénotype
Couple perturbation-observation faux positifs,
faux négatifs, traitement statistique,
formalisation de la conclusion
19
Integration of heterogeneous data

• Joint use of functional clues from a variety of
  experimental approaches to
• Validate the biological relevance of interactions
• Determine the function of proteins
• Validate targets in-silico
• Examples
• Interaction expression
• Interaction 3D structure
• Location expression
• Phylogenetic profiles domain fusion
• Recent problem, drug discovery efforts bottleneck
• Frontier for the bioinformatics community
• Technology normalization, formats, ontologies
• Science automate (some) biological reasoning ?

20
Evaluating pathways representations

• Vincent Schächter, Hybrigenics, Paris
• Aviv Regev, Tel-Aviv University
• BioPathways Formalisms Workgroup

21
Evaluation scope untangling the web...

• Large body of literature, focusing on different
  biological phenomena and different theoretical
  issues
• A typical article on pathways may include one or
  more of the following
• A data-model, describing (a fraction of) the
  pathway universe of discourse
• A formalism, used to describe the data-model and
  to express algorithms / functions
• Description of algorithms based on
  characteristics of both the formalism and the
  data model
• Description of implementations of data-storage
  functionalities and/or of some of the above
  algorithms

22
Excerpt from target evaluation list non DE
formalisms

• Petri nets (basic, hybrid, self-modifying,
  time-dependent, hierarchical, mobile)
• Process algebra (basic and stochastic
  pi-calculus)
• Markup languages (CellML and SBML)
• Biocalculus
• Regulatory grammars (Collado-Vides)
• Semiotes (Kazic)
• Statecharts (Kam, Holcombe)
• Boolean networks (basic, multi-level)
• Hierarchical networks (Bodnar)
• Neural networks (Mjolsness)
• Molecular graph reaction networks (McCaskill)
• Molecular interaction maps (Kohn)
• Electrical circuits (Keane)

23
Quelques exemples de représentations discrètes

• Modèles orienté-objet
• Requêtes sur tous types de réseaux
• Reconstruction, mais problème de l information
  incomplète
• Réseaux booléens
• Simulation qualitative, reconstruction à partir
  de données d expression
• Appliqué aux réseaux de régulation
• Réseaux de Petri
• Simulation qualitative plus fine, analyse
  formelle du comportement
• Appliqué aux réseaux de régulation
• Application possible aux réseaux métaboliques et
  signalisation avec extensions (self-modifying PN,
  Hybrid PN)
• Algèbres de processus
• Simulation, analyse formelle, reconstruction
• Appliqué aux réseaux de signalisation et de
  régulation (métabolisme avec extension
  stochastiques

24
The position of formalisms in the context of
pathways informatics

• Pathway construction
• Pathway generation
• Pathway selection

• Dynamics
• Simulation
• Analysis

Data storage retrieval Query language
Supports
Supports
Supports
Construction-oriented formalism data-model
Dynamics-oriented formalism data-model
Database-oriented formalism data-model
Expresses
Expresses
Expresses

• Core Representation / Ontology
• Biological scope
• Formal expressiveness

25
Evaluate and compare a modular approach

• Evaluate expressiveness/ease of use of
  representation relatively to specific
  goals/functionalities
• Compare representations in the categories for
  which they were designed
• Reduce each category to a set of evaluation items
  that can be rated and compared as objectively as
  possible

26
Core representation / Ontology

• Conceptual structure of the universe of
  discourse (abstract and concrete entities,
  relations, hierarchies...)
• Constrains scope of phenomena that can be
  described, and thus queried, analyzed,
  reconstructed, and queried.
• Often implicit in a given pathway representation
  need to extract...

• Possible evaluation schemes
• 1. Compare features of ontology
• 2. Expressiveness benchmark set of biological
  situations
• 3. Translation of data models into common
  formalisms comparison

How do you represent gene A inhibits gene B
in your data model ?
27
Conceptual Model Biological Scope Evaluation
Items
28
Core Representation Formal Expressiveness
Evaluation Items
29
Data Storage and Retrieval

• Storage and retrieval of data
  database-related functionalities
• Extremes relational or OO models vs, e.g., most
  simulation-oriented formalisms...
• A data-retrieval oriented formalism can be used
   below  other formalisms
• Query language
• Retrieve information within a structured,
  homogeneous, compositional framework
• Shifting boundary with analysis and
  reconstruction algorithms

• Evaluation items / sub-categories
• Robust database implementation issue
• Query language ease of use
• Query language expressiveness
• Limited by formalism and ontology expressiveness

30
Pathway reconstruction

• Construction/prediction of pathways in given
  biological environment (organism, tissue,
  condition, location) from a combination of
• experimental data
• fully instantiated pathway information,
• partially instantiated (or incomplete) pathway
  data, such as interaction data
• Special cases reverse engineering, pathway
  inference

• Evaluation items / sub-categories
• Input data types
• Pathway generation algorithm
• Pathway selection algorithm
• Pathway fitness function
• Pathway similarity/homology measure
• Interactive validation ?

31
Dynamics

• Study of network dynamics (regulatory networks,
  ST, MP)
• Simulation runs
• Analysis of dynamic behavior

• Evaluation items / sub-categories
• States nature, expressiveness, level of detail
  vs available data
• Evolution rules / Reaction model rule,
  implementation
• Time continuous/discrete, synchronous/asynchrono
  us updates
• Space continuous/discrete, topology, resolution
• Analysis
• Scope state reachability, liveness of
  transitions, substance flow...
• Formal methods available
• Comparative power
• Limited to steady state ?

32
Methodology what do we evaluate ?
Queries
Reconstruction
Simulation
Supports
Formalism
Evaluation targets
Describes
Data-model
Translation into common ontology description
language ?
Ontology