• Like
Bm Systems Scientific Epa Conference Heuristic Mathematic Concepts Synergies And Achievements Cns 2011
Upcoming SlideShare
Loading in...5

Thanks for flagging this SlideShare!

Oops! An error has occurred.

Bm Systems Scientific Epa Conference Heuristic Mathematic Concepts Synergies And Achievements Cns 2011


IMPORTANT: If you want to get a clear review of the Differences & Complementarities Between « Heuristic » and « Mathematical » approaches, we invite you to download our presentation given during …

IMPORTANT: If you want to get a clear review of the Differences & Complementarities Between « Heuristic » and « Mathematical » approaches, we invite you to download our presentation given during the EPA (European Psychiatric Association) conference in 2011 that is now utilized in training programs.

  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads


Total Views
On SlideShare
From Embeds
Number of Embeds



Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

    No notes for slide


  • 1. Methodology of Mapping in Integrative Molecular Biology. the Differences & Complementarities Between « Heuristic » and « Mathematical » approaches Concepts & Examples Dr. François Iris: Founder & CSO BMSystems To contact us: manuel.gea@bmsystems.net 19th European Congress of Psychiatry – EPA 2011 Vienna Austria, March 12-15 2011© BMSystems 2011; www.bmsystems.net
  • 2. The problem of Complex System Analysis and Biological Modelling. If you dream to create the first Be sure to use the appropriate operational bird model… modeling concepts & tools. If not… … a “basic” living Complex …you get a Complicated “Cartesian” system that not only flies… system. It does fly, but… it will never lay eggs! The challenge is clearly not a question of technologies only© BMSystems 2011; www.bmsystems.net
  • 3. Analysing & integrating data. Literature sources & text mining Data flatfile Analytical Web resources Visualisation tools© BMSystems 2011; www.bmsystems.net Networks building
  • 4. Currently, some 317 web resources (databases) provide access to• Thousands ofpathways and networks,documenting.• Millions ofinteractions betweenproteins, genes, smallmolecules.(http://pathguide.org/) © BMSystems 2011; www.bmsystems.net
  • 5. The challenge is to use as many of these biological databases as possible, CONCURRENTLY!© BMSystems 2011; www.bmsystems.net
  • 6. In practical terms, it means extracting as much information as possible from archived records.© BMSystems 2011; www.bmsystems.net
  • 7. The Classical Process The classical data mining approach pre-supposes that • All reported information is a priori valid, and • There are well defined rules for expressing relationships between components and they are practically always obeyed by authors. • Distance of co-occurrence : N phr (G) . N phr (G ) logHence the systematic N phr (G et G ) 2 application of mining algorithms • Functional relationships: such as N phr (G1 , G 2 , R) to all ‘omics data, from transcriptomics to proteomics & metabolomics© BMSystems 2011; www.bmsystems.net
  • 8. Complex System Analysis and Biological Modelling It is first and foremost integrating masses of information In the bio-medical realm, the relevant information touches on every biological domain, from anatomy to molecular biology and structural biophysics, spanning a multitude of functional contexts, corresponding to an enormous complexity. The available information is necessarily ALWAYS - incomplete to an unknown extent; - biased to an unknown extent; and - erroneous to an unknown extent. All that goes by the name « Information » is not necessarily Useful and/or Utilisable© BMSystems 2011; www.bmsystems.net
  • 9. The Classical Process Séquences 1, 2, 3 …n Data of interest Data acquisition Injection SPAD components Identification OMIM Nucleic acids Proteines, etc.. filters filtres indexation indexation DB IdentifiedMEDLine Indexed components Text mining, Manipulation, C1,C2,C3 … Cn Etc.. Exploitation What constitutes a GOOD filter? What constitutes a GOOD indexation strategy? Accumulation ALL information entered into the DB is ALWAYS of inconsistencies biased, incomplet, erroneous, etc…© BMSystems 2011; www.bmsystems.net
  • 10. This, naturally, generates coherence issues affecting the entire analytical chain of events, irrespective of the visualisation and manipulation tools finally used! Integrator© BMSystems 2011; www.bmsystems.net
  • 11. So we not only end-up facing ever more intricate networks which require massive manual curation to become utilisable. IP= Building blocks; M = distance of association; N= interactions© BMSystems 2011; www.bmsystems.net
  • 12. This also results in a highly misleading vision of protein interactions & networks. How can a single hub protein bind so many different partners? The problem is largely non-existent and resides in the construction and the representation of protein interaction networks. Proteins derived from a single gene, even if different, are clustered in maps into a single node. This leads to the impression that a single protein binds to a very large number of partners. In reality, it does not. Protein networks reflect confusions involving combinations of multiple distinct conformations of a same protein as well as functional plasticity addressing distinct proteins encoded by one gene.© BMSystems 2011; www.bmsystems.net
  • 13. One protein = Functional plasticity Post-transcriptional modification of Actin Actin His73 methylation = ATP exchange rate increased, ATP hydrolysis and phosphate release prior to and independent of filament formation while polymer formation delayed. Cytoskeleton dynamics significantly altered. DCD44 glycosylation patterns = drastically different effects on regulation of immune cells.© BMSystems 2011; www.bmsystems.net
  • 14. As a result, when it comes to reallycomplex systems and poorly definedphysiological processes, such asthose associated with functionswithin the CNS and disordersthereof, Bayesian approaches leadto nearly intractable difficulties.Mathematical models are remarkable validation/fine-tuning tools when applied to well defined processes. They are inadequate discovery tools when applied to poorly understood multicellular processes. © BMSystems 2011; www.bmsystems.net
  • 15. An example of what must be avoided: the systems biology of Creutzfeld-Jakob Disease.This Bayesian model waspublished in 2009(Hwang D et al. Mol Syst Biol. 5:252.PMID: 19308092) Constructed from a list of 333 shared DEGs and protein–protein interactions information from novel targeted experiments and public databases, it describes the networks that are potentially involved in the activation of microglia and astrocytes.© BMSystems 2011; www.bmsystems.net
  • 16. Although it certainly does lead to sub-networks that do make some sense, thesecannot 1) be distinguished from Alzheimer’s Disease, also characterised by astrogliosis, nor 2) begin to explain why neurons are killed in the CNS and never in the spinal chord, nor 3) why the disease is characterised by a spongiosis absent in other neurodegenerative disorders, nor 4) why the disease progresses silently for many years followed by a sudden very rapid clinical progression. Mathematical models are remarkable validation/fine-tuning tools when applied to well defined processes. They are inadequate discovery tools when applied to poorly understood multicellular processes. © BMSystems 2011; www.bmsystems.net
  • 17. As compared to the heuristic CJD model finalised in early 2008. Which predicts and explains the pathological mechanisms at both molecular and physiological levels. Source: BMSystems & CEA© BMSystems 2011; www.bmsystems.net
  • 18. The differences between « heuristic » and « mathematical » approaches. Heuristics: A problems solving approach evaluating each step in a process, searching for satisfactory solutions rather than for optimal solutions, using all available qualitative information instead of quantitative information. Thus, Heuristic modelling starts from accumulated information to produce a model capable of describing the mechanisms that generated the observed outcome / data and predict their modifications associated with a different outcome; It plays the role of an architect. While mathematical (Bayesian) modelling starts from quantitative data to produce models capable of reiterating this data and predict the outcome of a different experimental paradigm. It plays the role of an engineer. Hence Far from being incompatible, these two approaches are complementary.© BMSystems 2011; www.bmsystems.net
  • 19. The heuristic analytical process must follow a « relativistic » approach. Within this framework, Non-linearity, Irrelevance, Wear, Relative weights & Contexts are key concepts.© BMSystems 2011; www.bmsystems.net
  • 20. And the « relativistic » approach leads to a very simple set of rules. Events tell contexts how to evolve Contexts tell components how to behave Components tell events how to arise Analyses in terms of biological components and functions are now IRRELEVANT. EVENT-DRIVEN analytical approaches become necessary. This, in turn, imposes analytical procedures based upon the negative selection of working hypotheses.© BMSystems 2011; www.bmsystems.net
  • 21. The CADI™ Integration & Modelling Process. Séquences 1, 2, 3 …n Data acquisition Data of Interest components Identification Nucleic acids Proteines, etc.. Injection SPAD Identified components OMIM Filters indexing Indexed C1,C2,C3 … Cn MEDLine DB Index visualisation manipulation The DB contents are Negative Selection analytically of hypotheses IRRELEVANT© BMSystems 2011; www.bmsystems.net
  • 22. To arrive at a model that can be biologically challenged! Nucleic Data acids acquisition Proteins Etc… Data analysis Experimental verification Biological Validation Identified biological Hypothetic events Physiological Mechanism Literature Database Searching Biological Modeling |||||||| |||| |||| |||| |||| |||||||| Production |||| of working Components hypotheses Interactions maps Destructive hypotheses testing© BMSystems 2011;www.bmsystems.net
  • 23. The heuristic model of CJD finalised in early 2008. It predicts and explains the pathological mechanisms at both molecular and physiological levels . Source: BMSys & CEA© BMSystems 2011; www.bmsystems.net
  • 24. The main driving mechanism. Il-1b-mediated signalling inhippocampus astrocytesThe pathways through which chronicneuronal stress signalling and concurrentglial pro-inflammatory responses lead toreactive astrocyte activation (GFAP + Vim)associated with cytoskeletonreorganisation (ezrin). This leads to amajor switch in Cx targeting &distribution, resulting in the formation of asyncythium with massively altereddiffusive properties and neurotrophicfunctions. © BMSystems 2011; www.bmsystems.net
  • 25. Practical consequences.One of the role of connexins is to dampen neuronal synchronisation. Non infected Mice AntiCX 0,75 Non infected Mice NaCl Healthy mice 0,55 0,35 0,15 -0,05 Hz NaCl -0,25 -0,45 Anticonnexin -0,65 -0,85 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 In healthy animals, pharmacological blockade of Cx activity results in quantitative EEG patterns closely resembling an epileptic crisis (frequency range-specific hyper-synchronisations). Hence, in cases where the aim of treatment is to affect synchronisation, the adjunction of Cx modulators could potentiate current drugs therapeutic effects (allow to decrease doses while preserving activity) thereby reducing unwanted side effects. A new model was constructed and the predicted approach tested in vivo.© BMSystems 2011; www.bmsystems.net
  • 26. Effects Cx modulator adjunction to Paroxetin treatment. 1st hour 2nd hour Paroxetin 1 mg/kg ( 2 to 5-fold below usual doses ) Paroxetin 0.5 mg/kg + Cx modulator 0.4 mg/kgSimilar boosting effects of Cx modulators are seen withClozapin; Modafinil, Diazepam; Venlafaxin; Escitaloprame; Bupropion; Sertralin; etc. The dose-reducing effects are in the order of 5 to 20-fold (depending on the therapeutic molecule) without any loss of therapeutic effects. NB: one of the molecules utilised as Cx modulator is already on the market for a completely different indication (immunology) and is used here at 1/50th of its approved dosage (the predicted dose-dependent effects were observed).© BMSystems 2011; www.bmsystems.net
  • 27. The net results. CJD is not a neurological disease stricto sensus. It is a disease that primarily affects astrocytes structures and functions which, over time, lethally affects healthy glial & neuronal cells through « bystander effects », leading to widespread CNS disorganisation and functional failure. But this model also provides an understanding of key mechanisms associated with psychiatric disorders. An entirely new approach for their effective treatment was designed, tested in vivo and validated. Patent covering novel therapeutics for cognitive disorder (CEA/BMSys). PCT: WO 2010/29131 A1 (« class » therapeutics patent) • Creation of a spin-off under process (TheraNexus) Neither of which have much to do with CJD per se… This work received a Bio-IT World « Best Practices » award from the Cambridge HealthTech Institute (USA). AND Was selected as 1 of the 3 pan-European « state of the art examples of systems biology approaches of benefit to medicine » by the European Commission’s DG Research, Directorate of Health (June 2010).© BMSystems 2011; www.bmsystems.net
  • 28. Nevertheless, it MUST be remembered that Models are AIDS to thought NOT a replacement for it.© BMSystems 2011; www.bmsystems.net
  • 29. Thank you to Bio-informaticsThe “Engineering” Teams at BM-Systems, directed by Paul-Henri Lampe, Pablo Santamaria & Manuel Gea BiologistsThe “Integrative Biology” Team at BM-Systems, directed by Dr Francois Iris. and, most of all Academic collaboratorsThe CEA Prionics group, European Centre of Excellence, directed by Prof. Jean-Philippe Deslys & Dr. Franck Mouthon. ..and to you for your attention. © BMSystems 2011; www.bmsystems.net
  • 30. Questions To contact us: manuel.gea@bmsystems.net© BMSystems 2011; www.bmsystems.net