redefining social relevance of research through participatory approaches

September 25, 2014

In the recent years, mostly due to the economical crisis and evolution in science policies, the pressure on scientific research to be relevant for society is increasing (see picture from Utrecht). Current definition of relevance is usually restricted to economical outcomes such as technological and medical applications. Many scientists disclaim the pressure for relevance on their shoulder, arguing that serendipity has always been a driver in scientific discoveries. Expected relevance of research is therefore almost impossible, as genuine innovation cannot be predicted. Without entering this debate, which could be the topic of an entire blog-post, I propose that the current definition of relevance needs to be widen and take into account social relevance. I have recently take part in a new involvement of citizen in research (http://www.joursavenir.org/ncs/en), which I believe can help to define social relevance of research.

What is a socially relevant research? This question needs to be answered by Society at large but today we lack the tools to enable it. Researchers have been used for so long to decide what is a relevant scientific question among themselves. They dislike situations where they are evaluated by Society, arguing that most people don’t have enough knowledge to judge. Positioning people as evaluators of research generates even more divisions between scientists and non-scientists. Instead people can be involved since the start of the research, that is to say they can co-define the research questions with the scientists (and also perform the research but I’m not discussing this part). In such a co-definition process, the question is neither the one of scientists approved by citizens, neither the one of citizens received by scientists, but it emerges from a shared interest. A research being co-defined with citizens will therefore increase its relevance by directly being in dialogue with groups of people who benefit from the answers in using the research results. Some research in agriculture has already shown that it is feasible to involve people in the definition of the research question (ref). I believe this can be extended to other types of research, including very fundamental fields.

To achieve this, new tools that involve groups of people into the construction of the scientific research need to be invented. Citizen science research has already developed protocols that involve citizens in fundamental research but the involvement is restricted to data analysis (ref). With Atelier des Jours à Venir, we have started to create and experiment new type of tools in the context of the New Patrons of Science’s project (http://www.joursavenir.org/ncs/en), funded by Fondation de France. This project identifies groups of citizens that have a common interest and have them co-design and perform a research with researchers.

We have chosen to start experimenting these tools together with people that are far away from the corridors of power where science policy is decided. For our tools to be valid, it has to work with the people who are usually not represented when the relevance of research is decided. If we want to achieve more social relevance to research, we need to get a diverse representation of the interests of people. Although it may seem a challenge, we have already shown that such public can ask genuine scientific questions (http://joursavenir.wordpress.com/2014/08/18/science-magazine-aug-15th-2014-kids-questions-transcend-conflicts/). The future will tell us in which context it is possible to involve groups of people in the design of a research questions and we will inform you on our progress.

To conclude, as a researcher, I would like to say that working on research questions that are co-defined with citizens not only makes it easy to fill in the relevance part of your research grant. The process of co-construction can make your research project evolve towards more society-relevant issues. This strengthens your motivation by connecting your work to one of the reasons we are involved in research: contributing to society.

This article has been co-published on both the immune modeling and the atelier des jours à venir blogs.

 

Advertisements

On the difficulties to model CFSE data

December 3, 2013
Last summer, with Fanny Moini, a brilliant master student, we have been
working on modeling CFSE data. In brief, CFSE is
a dye that is diluted by two fold following each cell division. One can
record the number of cells in a particular division category at a
particular time and offer in theory a rich source of information to
mathematically model cell dynamics. It has been extensively used to
model the dynamic of lymphocyte expansion (1,2). Several competitive
mathematical model offer to describe CFSE data. For a new comer in the
field, it is difficult not getting lost among those models and decide
which model is the most appropriate for particular CFSE data.  In this
post, I wish to share some of our conclusions that could be useful to
new comers and announce the online release of open source code to model
CFSE data.

Since 1999, several mathematical models, from a relatively simple
precursors evolution analysis (3) to more complex probability theories
such as branching processes (most recent review see 2) have been
developed to model CFSE data. Those models differ in their number of
parameters, their assumptions about death and proliferation but also in
their implementation, which makes their comparison difficult. We first
build a chart to compare the three most commonly used model (see figure
below, 4,5 and 6), in hope it will help people visualize the differences.

It would be interesting to be able to compare these model on the same
set of data using the same uniform procedure. For this purpose, we have
made two of the current models implementations available (The
Smith-Martin implemented as in 5 and the Cyton model implemented as in
6). We have added  additional useful features (such as visual outcomes
and Akaike Information Criterion calculation) and you can find the
scripts here.

Our conclusion after playing with a test data set where several time
point have been recorded are disappointing and join other people's
conclusion (1,2). Although CFSE data contains much information about the
dynamic of the cells, it is not enough to reliably model the dynamics.
All the models suffer from parameter compensation for example cel death
and proliferation rate can compensate each other, despite the large
amount of data.  Additive information obtained with another experimental
technique for example about the death rate or empirical assumptions to
reduce the number of parameters seems to be essential.

references:
1. Callard, R., and Hodgkin, P. (2007). Modeling T- and B-cell growth
and differentiation. Immunol. Rev. 216, 119–129.
2. De Boer, R.J., and Perelson, A.S. (2013). Quantifying T lymphocyte
turnover. J. Theor. Biol. 327, 45–87
3. Nordon, R.E., Nakamura, M., Ramirez, C., and Odell, R. (1999).
Analysis of growth kinetics by division
tracking. Immunol. Cell Biol. 77, 523–529.
4. Hawkins, E.D., Turner, M.L., Dowling, M.R., van Gend, C., and
Hodgkin, P.D. (2007). A model of immune
regulation as a consequence of randomized lymphocyte division and death
times. Proc. Natl. Acad. Sci. U.S.A. 104, 5032–5037.
5. Ganusov, V.V., Milutinović, D., and De Boer, R.J. (2007). IL-2
regulates expansion of CD4+ T cell
populations by affecting cell death: insights from modeling CFSE data.
J. Immunol. 179, 950–957.
6. Zilman, A., Ganusov, V.V., and Perelson, A.S. (2010). Stochastic
models of lymphocyte proliferation and death. PLoS ONE 5.

Quorum sensing effect in the immune system

March 26, 2013

I would like to point out one of my recent publication about quorum sensing in the immune system. I explored the hypothesis of existence of quorum sensing like mechanism in the immune system. This work aim to motivate immunologist to think more of population dynamic in the immune system. It was done during the year after my PhD during my research assitant position at College de France.

It was not easy to publish a hypothesis paper like this one has a young researcher. Even if the work is not completly mature, it is nice that there is some journal that accept to publish such work.

References:

Does a quorum sensing mechanism direct the behavior of immune cells? Comptes Rendus Biologies. Volume 336, issue 1, January 2013, Pages 13-16

link to the publication on the publisher’s website

link to the unedited version on my website


L’hématopoïèse à l’échelle unicellulaire

February 13, 2013

I am honoured to be the next guest of the philosophy and immunology seminar series (that I created 3 years ago before leaving to the Netherlands). The session will be done in French, here is the information about the seminar.

La prochaine séance du séminaire Philosophie & immunologie aura lieu le 21 février 2013, 17h-18h30 (merci de bien noter l’horaire inhabituel), à l’IHPST :
L’hématopoïèse à l’échelle unicellulaire : changer d’échelle et de regard sur un processus biologique grâce au code-barre cellulaire
Résumé :
L’hématopoïèse est définie comme le processus de différentiation qui conduit à la production des cellules sanguines à partir des cellules souches hématopoïétiques. Cette différentiation a été étudiée sur des populations de cellules et est classiquement représentée sous forme d’un arbre de différentiation. Les branches de l’arbre conduisent à la production des cellules immunitaires et des autres cellules sanguines par diversification de leurs précurseurs. Deux lignées principales divergent à la base de cet arbre : une lignée myéloïde et une lignée lymphoïde.
L’avancée technologique des code-barres cellulaires, nous permet d’étudier ces processus de différentiation non plus simplement à un niveau global, mais à l’échelle des cellules individuelles. Or, nos récents résultats expérimentaux remettent en cause l’arbre de différentiation tel qu’il a été accepté au cours de ces dix dernières années. Cet exposé se propose de discuter en quoi un changement d’échelle technique et conceptuelle dans l’étude des objets biologiques amène à repenser en profondeur un processus biologique tel que la différentiation. En s’appuyant sur une définition généalogique de la lignée plutôt que moléculaire ou génétique, la technique des code-barres cellulaires révèle ainsi l’existence d’une troisième lignée, producteur des cellules dendritiques. Par ailleurs nos résultats couplés à de la modélisation mathématique suggèrent que le processus de différentiation ne suivrait pas des voies uniques (« les branches de l’arbre ») mais emprunterait des voies multiples. Nous discuterons des implications théoriques de ces résultats en soulignant la nécessité d’une recherche de nouvelles analogies pour exprimer de façon appropriée un nouveau regard sur ces processus.
Lectures conseillées :
–  La technique des codes barres cellulaires : Nat Rev Immunol. 2010 Sep;10(9):621-31. Mapping the life histories of T cells. Schumacher TN, Gerlach C, van Heijst JW.
–  Histoire de la découverte des cellules souches hématopoïétiques :
Stud Hist Philos Biol Biomed Sci. 2007 Mar;38(1):217-37. Epub 2007 Feb 12.The search for the hematopoietic stem cell: social interaction and epistemic success in immunology. Fagan MB.

War and Police metaphors in immunology : updated version

November 21, 2012

I am concerned about the use of the war and police metaphors in immunology. Since these metaphors are so prevalent in immunology, I wonder if it prevents us from approaching ideas in a new way.  The war and police metaphors are so engrained in our education that I find it very difficult to think outside of these specific boxes; general communication about immunology always uses these metaphors. I have the feeling that I am up against a brick wall when I try to think differently. I would like to explore other metaphors.

I find the war and police metaphors too extreme because they always use repressive and aggressive vocabulary (attack, kill, repress…etc). If you try to avoid this extremism, it is tempting to go to the other extreme: a peace and love metaphor (ref 1). However this could be an object of ridicule and is just as extreme as the war and police metaphors. I have chosen to take a different angle here.

We first need to understand the limitations of the war and police metaphors.  Leaving aside the anthromorphism issue of the war and police metaphors (I will save this discussion for another time), my major opposition to these metaphors is based on the underlying conservative vision of human society that is extended to the human body. You could argue that I should not mix science metaphors and political visions. I argue that being aware of these influences brings us one step closer to objectivity. Because I don’t necessarily share this conservative vision, I notice that using these metaphors can be a problem. I would like to explore another politically bias metaphor in immunology.

My vision of the immune system is of a system that maintains homeostasis. The immune system would then play a role in the maintenance of the equilibrium of the organism. Michod has proposed a role for the immune system in the emergence of multicellularity (ref 2). I would like to go a little bit further and propose that the immune system plays a role in the maintenance of this multicellularity.

This maintenance can be seen as human social cohesion systems, like social security or justice. Originally the concept of immunity comes from the need to offer juridical protection for those who work for the organization of society (ref 3). Using a social cohesion metaphor connects us to the etymology of the world immunology. The war and police metaphors are now welcome as part of a bigger metaphor of human social cohesion systems. We also use of less aggressive vocabulary such as equilibrate, equity, help etc…

The invention of this new metaphor is already giving me some new ideas on the organization of the immune system. We have a hard time defining the actor of the immune system. For example, are the epithelial cells parts of the immune system? Could some local actors such as library or cafe, which are not part of the central social system, achieve social cohesion? If so, some actors of the immune system can also maintain homeostasis at the local level but still play a role in the social cohesion. Therefore they should be included in the definition of the immune system.

Of course, this is a first idea that needs to be further developed. I hope you will find it interesting. I am looking foward to hearing your suggestions and to dicussing it with you.

Refecences:

1. Ed Cohen. Metaphorical Immunity.  Literature and Medicine 22. 2003.

2. Richard Michod & Denis Roze, Cooperation and conflict in the evolution of multicellularity, Heredity, 2001.

3. Peter Sloterdijk personal communication at Paris Senat 15/06/2009.

 


High throughput data and multidimensional analysis II

June 6, 2012

Continuing on multidimensional analysis of high throughput data, I would like to share my surprise on how close is the limit of knowledge in that subject applied to biology! I have been analyzing for 9 months immunological high throughput data that are not gene expression, starting from knowing very little about bioinformatics. I arrived very rapidly to the limit of our knowledge, that is to say to an open question for the specialists in the field! Let’s take one concrete example.

If I take hierarchical clustering as an example, choosing the method (ie the algorithm and the distance method to apply) is not trivial. There is no standardized method available. Usually the answer given is to choose a common method or the one that looks good. I don’t know what is a “good looking” data (unlike for men!). Moreover there is biological differences in the interpretation of data depending on which method you choose. For example it is not the same if 2 different groups of cells cluster together or not. I have used some tricks to determine which method was appropriate, which consist of putting some control samples in the middle of my real sample and use them to choose the method. I have also found that correlations such as Pearson are really sensible to zeros (which tend to be correlated but this is not what you often want). There is often a lot of zeros in biology. These zeros can be meaningful or not. For example if a patient doesn’t respond to a treatment or has no more tumor, well it is meaningful data, although this can be recorded through a zero value data! Looking in the literature, I haven’t found a satisfying solution to the problem; it seems to still be an open question in the field. In my research, I have finally chosen to use Euclidean distance but there is also some limitations there. In consequence you have to be really careful with your conclusions.

Now, when I listen to a talk of a biologist talking about high throughput data, I am often not convinced by the result because the choices for the analysis are not explained. You have no idea on how robust is the conclusion toward the analysis.  Reading articles, you often have access to more information on the analysis. Often there is too little explanation on the rational of the choice of analytical tools. It is a pity because it is useful information to judge the quality of the results.


High throughput data and multidimensional analysis I

May 22, 2012

Nowadays it is quite fancy to have high throughput data or some experimental data associated with a mathematical model. These types of work often publish quite well, in high impact factor journals. It requires a strong collaboration between experimentalists and bioinformaticians or theoreticians. One can’t work without the other one, experimentalists often can’t analyze alone the data because they are multidimensional and difficult to interpret without statistical models, and vice versa theoreticians need obviously data, while they lack the knowledge about how to acquire quality data.

What is seldom discussed within the community is how to manage well the collaboration. I would like to talk about some difficulties that are faced in the field. To be concrete I will address the issue of authorship, because it is an essential part of the research activity (sometimes too much!).  The person who performs the majority of the experiments, design the study and interpret the data is usually the first author in classical paper. In the case of high throughput project, the interpretation of the data and even the design of the study are closely related to the bioinformatical analysis. But often the bioinformaticians is still seen as a technical help and thus is offered a position behind the experimentalist in the paper.. The number of people with a mixed background increase more and more, bioinformaticians are now well trained in biology and their work is becoming more than just providing tools for analysis. They really have an impact on the interpretation of the data .Experimental labs often underestimate bioinformatics. I did have the same position before putting a bit of hand in bioinformatics.  But on the other hand the experimentalist also deserve credit because doing experiment is long, complicated and painful!

How can we solve the problem of authorship? First joint authorship is often proposed as a solution. As the number of joint authorship is increasing that’s maybe a good solution but we all know that it is not seen as the same to be the first first author or second first author (I know that it sounds crazy…).  Another option is to split the publication of the methods and the experimental data when possible.  I don’t have better solutions to propose. I think this topic still has to be discussed more broadly within the community of scientists

To end this post, one general comment about the multidimensional analysis. I was surprised how little tools we have to visualize multidimensional experiments. There is urgent need of more input of people working on visualization tools to develop tools for biologists! It is ridiculous to have more tools to analyze facebook network than biological data! And what about the authorship of the people working on visualization?