FAQs

DYNAMIC MODELING :

What does Helios Biosciences model?
We model the dynamics of intra-cellular signal transduction pathways in response to various stimuli. The models are at the molecular level. The stimuli may be either physiological ones (ex.: stimulation of T-lymphocytes by antigens) or pathological ones (ex.: glutamatergic hyper-stimulation of neurons). Most drugs used in human clinics target signal transduction pathways.

Does dynamic modeling work?
Yes it does. Our dynamic models accurately account for complex, non-linear and non-additive molecular responses, in large signal transduction pathways involving several hundred molecules (see our T-lymphocyte experiments described in our brochure. Even on long time-courses (ex.: 48hours) the overall fit between the models and the experimental data is above 80% which is within the experimental measurement error. We always simultaneously model more than one stimulus / cellular response to produce stringent models.

Is dynamic modeling really predictive?
Yes it is. Statistical analyses show that the predictive power is highly significant (p<10^-10). Simulations based on our human T-lymphocyte response model have identified a relevant and limited panel of drug targets. The benchmark analysis shows that one third of the selected targets are already experimentally validated (of which some are already targeted by drugs used in human clinics). The remaining ones will now be tested.
This is to be compared to the high attrition rate of drug targeting by other existing methods (>90%).

Which diseases may be addressed through dynamic modeling?
Helios Biosciences main focuses are neurology, immunology and cancerology. However, any pathology may be addressed providing that biological samples relevant to it may be produced during a time-course. Typical examples are the time course of the onset of a disease, or time course of the response to a drug…). Cellular models, animal models or human samples may be used.

Which kinds of data are required to build a dynamic model?
Dynamic models integrate two kinds of data:
A description of the molecular interaction pathway of interest, and
At least one time-course experiment describing the expression profiles of the molecules within this pathway.
We provide the molecular interaction pathways through our Simpathway database. We can also take charge of producing the experimental data.

Is it always possible to build a dynamic model?
The ability to build a dynamic model of a physiological / physiopathological phenomenon is constrained by the availability of data on this phenomenon. In most cases the data have to be produced, and the technical feasibility of producing relevant data is acquired.

Why use dynamic modeling instead of other drug target identification methods such as :
Expression profile statistical analysis (gene clustering, expression profile mapping on molecular interaction pathways)?
Our extensive benchmark analyses within the framework of the T-lymphocyte response show that:
Gene clustering or variation-of-expression rates are not efficient ways to select drug targets, because the expression profiles of validated drug targets do not statistically differ from the other molecule ones within a given pathway. This may partially explain the high current attrition rate in drug targeting (selection of inefficient targets).
Expression profile mapping on molecular interaction pathways only identifies the highly connected molecules (‘hubs’) which expression strongly changes. Acting on such molecules is a good way to produce major side effects, and as a matter of fact most drugs used in human clinics do not target such molecules. This may also partially explain the high current attrition rate in drug targeting (selection of too much toxic targets).
The drug targets identified through our dynamic modeling integrated approach include non-hub & non-strongly modulated molecules which are experimentally validated targets.
Dynamic modeling also provides the ability to systematically assess associations of effects (co-targeting) which is not feasible through other techniques.

Systematic inhibition of all molecules within a given molecule interaction pathway (RNAi and other)?
Such strategies are ‘systematic’ only at a small scale which is a major limitation (testing of a few to a few dozens molecules). They also require the previous definition and delimitation of the pathway. Our expertise in human pathway building unambiguously shows that to be coherent, a pathway sustaining even a limited physiological function (ex.: signal transduction from a ligand / receptor couple) systematically includes more than 100 genes / proteins.
Only dynamic models provide really systematic analyses which means testing hundreds of molecules.
Dynamic modeling also provides the ability to systematically assess associations of effects (co-targeting) which is not feasible through other techniques.

Why search for new drug targets?
Current treatments are insufficient in many diseases. Moreover, there is no efficient treatment at all for large fields of human pathology such as most neurodegenerative diseases and many metastatic cancers (see Prostate Cancer on our website). In such cases there is no relevant clinically-tested molecular hypothesis to support new drug development. Generating relevant molecular therapeutic hypotheses is thus crucial. This is what dynamic modeling is made for.

Which are the specificities of Helios Biosciences modeling technology?
We model large-scale signal transduction pathways, integrating non-linear non-additive effects, with convergences / divergences / feedbacks in signal propagation. This enables us to build models that stand close to what really happens in a cell, and to avoid being limited to over-simplifications.

Is a dynamic model that has been developed to address a given pathology transposable to other diseases?
We build one model to address one specific question in one disease. Addressing another pathology requires building another model, based on other data. However, if the diseases have many similarities, the building of a new model will be simplified, and will mostly depend on the generation of additional experimental data. For example, a model of the response of prostate cancer cells to sexual steroids will be easy to adapt to another hormone-sensitive cancer.

Why did Helios Biosciences choose the T-lymphocyte response to bring its dynamic modeling proof of principle?
The T-lymphocyte paradigm is one of the best to establish a robust proof of principle, because of its very high complexity. If a modeling method works for T-Lymphocytes there are few doubts that it will perform well when applied to other biological questions.
It is also a paradigm that is relevant to human clinics, because it explores crucial issues with regard to inflammation, allergy, auto-immunity and graft tolerance. Last, the existence of already validated drug targets in this paradigm provided us with benchmark data to validate our method.
Complexity of the T-lymphocyte response: the response of T-cells to the presentation of an antigen includes either cell proliferation and terminal differentiation or cell death depending on the antigenic context and time after presentation. It is thus a highest complexity response, sustained by dynamic changes in the intermixed signal transduction pathways of cell differentiation (including specificity of lymphokine synthesis), cell cycle and apoptosis. Whatever the type of the response all these three pathways are simultaneously engaged. Hundreds of molecules, interlinked through thousands of interactions, are involved, with non-linear and non-additive signal redundancies, convergences, divergences and feedbacks. Moreover, the dynamics of the response are very heterogeneous at the molecular level, different molecules reacting differently with different time delays even in close topographic areas of the pathway. This is what we have modelled.

Is dynamic modeling suitable to predicting potential drug toxicity? Would it be possible to predict drug toxicity?
Though theoretically feasible, such an application would require the availability of reference data about drug toxicity, in an appropriate format. As we do not have access to such appropriate data, we cannot address this issue for the time being. Nevertheless, we are always interested in finding appropriate partners to address new application fields

FAQs

SIMPATHWAYS :

Which molecules are in the network? How are they selected?
We build a signal transduction network. The molecules are included based on their role in signal transduction. Most drugs used in human clinics act on signal transduction.
Any kind of molecule is included whenever it is relevant to signal transduction in the physiological function considered. Our network includes ligand/receptor couples, linkers, kinases, phosphatases, proteases, transcription factors/cofactors, enzymes synthesising second messengers and other kinds of molecules. The general organisation of the network is based on interconnected signal transduction loops. The molecules are selected based on their effect on cell phenotype/response to stimuli. See Smart Integrated Molecular Pathways on our website.

Is the network biologically relevant?
Yes it is. We used our network to model the dynamics of the T-lymphocyte response. This model provides relevant results. The network does account for more than 80% of its molecule expression changes during the T-lymphocyte response.

Which criteria are used to select the interactions between molecules?
Generally speaking, interactions are accepted only when multiple criteria corroborate their functional existence and effects.
Interactions are selected based on the availability of their rigorous description in one or more scientific publications.
The major criterion is the unambiguous and specific involvement of the interactions in defined signal transductions: impairing the interaction must have a phenotypic consequence.
Additional criteria are also required: the interaction is proven to be direct (in the case of signal transduction mediated through second messengers (cAMP etc.), one intermediate second messenger molecule is accepted), experimental data enable to unambiguously direct the interaction …

Example: a transcriptional interaction (transcription factor A regulates the expression of gene B) is included in the database if :
- A physically interacts with regulatory sequences of gene B, and
- the responsive element in the regulatory sequences of gene B is identified, and
- mutations in the responsive element of gene B or changes in the amount of A or mutations in the interacting motif of A change the expression of B. For this, the minimum experiments required are mutagenesis experiments with transfections of the B promoter with a reporter sequence in a whole cell system; in this case, indirect evidence of changes of the expression of B mRNA in a whole cell system is also required. Direct evidence of changes of the expression of B mRNA in a whole cell system is better.
- additional criteria may be also taken into account : the promoter of B includes a consensus sequence for the transcription factor A, knock-out data or gene expression data…
The existence of a consensus sequence for the transcription factor A in the promoter of B is not sufficient to accept the interaction, neither is the physical binding of A on the B promoter…