Travel grant reports 2009

Carlos-Eduardo Valencia-Alfonso, Netherlands Institute for Neuroscience, The Netherlands

Purpose

NWG-Course “Analysis and Models in Neurophysiology” at the Bernstein Center for Computational Neuroscience (BCCN) in Freiburg (Germany).

Scientific part

My main interest in participating in this course was mainly to learn a wider range of analysis possibilities for electrophysiological signals, as well as the software tools to apply them to our data.
There was an introduction on the first day with a presentation of the participants, tutors and speakers. In each the following days we had a lecture during the morning, accompanied by a practical session in the afternoon.
During the second day, we had the opportunity to learn about the basics principles of systems and signals. Here we focused on the mathematics underlying signal analysis, how these determine the type of variables measured and they relations. The content of this session included theory of systems, their linearity, transfer function and Fourier analysis, as well as correlation analysis and non-linear systems. In the practical session we solved exercises regarding measurements of impulse, sine-waves and stochastic signals using Mathematica software.
The third session summarized point processes and correlation measures. Here we had an overview of the electrical properties of neurons, synapses and neuronal interaction.  The practical session was again using Mathematica and involved exercises with simulated signals in order to compare Poisson, Renewal, Convolution and Gamma processes.
In the fourth day we reviewed spike train analysis and correlation measurements. Here we analyzed basic measurements like Peri-stimulus-Time Histograms, Spike-Trigger-Average, rate etc, and the principles of cross-correlation.  The practical work included exercises on Trial management of signals, their temporal structures, correlations and unitary event analysis in Matlab, using a toolbox developed by the speaker.
In the fifth day we were concerned with local field potential (LFP) and the analysis of neuronal population activity. We also learned how to conceptualize and observe LFPs and Multi-unitary activity dynamics and the extraction of heir singular characteristics. In the afternoon session we developed Matlab based exercises, also with a toolbox developed by the speaker (MEA tools 2.5). Here we learned how to review the raw data, clean noise, choose relevant analysis windows and calculate response rates for the different conditions of the experiment.
During the last day, we were introduced to FIND (Finding Information in Neural Data). FIND is a Matlab toolbox for electrophysiological recordings and network simulation environments. We worked with data from single- and multiple-electrode recordings testing different analysis and visualization methods.
In summary, it was a very rich experience, both academically and personal. I have now a wider comprehension of the dimensions and possibilities of signal analysis. I also gained knowledge, skills and technical resources for the handling of my own experimental data. At the same time, the use of these software tools will allow their developers to receive feedback and suggestions about it. We are currently testing these tools in our laboratory and agreed to stay in frequent communication with the Bernstein center in Freiburg. Finally, I would like to express my gratitude to the INCF. It is a wonderful labor to support the development of interdisciplinary work and to strength young researchers’ capabilities. This was a great opportunity to expand the range of my scientific analysis and technical competences.
 

 

Niraj Dudani, National Centre for Biological Sciences, Bangalore, India

Collaborating laboratory: Computational Biology and Neurocomputing group, School of Computer Science and Communication, Royal School of Technology

Purpose

The aim of the visit was to work along with the developers of MUSIC to make MUSIC and MOOSE compatible with each other. MOOSE will benefit from talking to other simulators and, at the same time, early adoption by notable simulators in the field will be an important factor in achieving success for the MUSIC standard.

Accomplishments

    • Got full match between the outputs generated by the GENESIS and MOOSE simulators, for the same model
    • Implemented the MUSIC interface in MOOSE
    • Enabled MOOSE to both send and receive spikes through MUSIC
    • The major advantage of this visit was that now the two involved groups are keen on continuing this interaction, for instance with porting MOOSE to the BlueGene architecture

 

Gaute Einevoll, Mathematical Sciences and Technology, Norwegian University of Life Sciences, Norway

Collaborating laboratory: INCF Polish Node, Nencki Institute of Experimental Biology in Warsaw, Poland

Purpose

The aim of this visit was to promote the further development of collaboration between the Polish and Norwegian INCF Nodes and in particular to plan the detailed arrangement of the 1st Polish-Norwegian Neuroinformatics Workshop held at Ski, Norway January 15th-16th, 2009. In addition, it was the opportunity to present research activities in computational neuroscience at the Norwegian University of Life Sciences to researchers at the Nencki Institute of Experimental Biology in Warsaw, Poland.

Accomplishments

  • Planned the LFP workshop, modelling and interpretation of extracellular field potentials, that was to be held in Ski, Norway. Meetings with Daniel Wojcik and his group at the Laboratory of the Visual System, as well as with other researchers at the Nencki Institute of Experimental Biology.
  • Gave seminar entitled  "How are  extracellularly potentials related to the underlying neural activity?"  attended by the researchers of the Nencki Institute of Experimenal Biology.

 

Andreas Klaus, Department of Neuroscience, Karolinska Institutet, Stockholm, Sweden

PRACE Winter School

Background

I am a PhD student at the Department of Neuroscience and my project involves large-scale modeling of neuronal networks (supervisor Jeanette Hellgren-Kotaleski, KTH). On the Blue Gene/L at KTH we simulate networks of biophysically detailed neurons with the neuron simulator PGENESIS.

Goals

The main aim of my participation in the PRACE Winter School was to learn more about existing supercomputer architectures, as well as existing and novel programming paradigms (e.g. PGAS, fragmented vs. global-view, HPCS). Furthermore, I wanted to learn about the performance measuring tools for large-scale parallel computers, which are essential for the programming and usage of optimal code on such systems. I originally intended to install PGENESIS on IBM's Blue Gene/P (BG/P) at Jülich as it was done last year for the Blue Gene/L system at KTH. However, due to the lack of time during the hand-on sessions and access to BG/P, I could not install PGENESIS on this architecture (see below).

Activities

The workshop was 4 full days. During lectures in the mornings advanced topics of parallel computing were introduced – including programming models (MPI, OpenMP, Hybrid), optimization techniques, PGAS programming with Unified Parallel C, next generation HPCS languages (Chapel), and the CELL system/programming. In the afternoons the lectures were complemented by hand-on sessions which covered all the above-mentioned topics. In the hand-on sessions we used for example the Huygens system in the Netherlands, which is a clustered SMP system (IBM Power6). In the afternoon of the last day we were more flexible in trying own examples/code and testing things of individual interest. I used this time to work in more detail with some performance measuring tools (see below).

Achievements

I learned about the programming, optimization and usage of peta-scale systems. In addition, I learned several performance measuring tools such as Scalasca, Kojak, Vampir and Marmot. In particular I worked with Scalasca and compiled and analyzed given examples and an own simulation. This knowledge will be of immediate use for my work with the Blue Gene/L system at KTH (large-scale modeling of realistic neuronal networks). The lack of time and access did not allow me to compile PGENESIS/MOOSE on the Blue Gene/P in Jülich. I will, however, supplement the PGENESIS package at http://software.incf.org, which is for the Blue Gene/L system at KTH, by some additional information about scaling and performance.

Outline

In near future we will use MOOSE, the successor of PGENESIS, and optionally some special-purpose code on supercomputers such as the Blue Gene/L. Since the up-scaling of neuronal networks is not straightforward (it for example depends on the connectivity between the neurons), knowledge about the underlying hardware architecture (i.e. their computing nodes and the interconnecting network) and about performance measuring tools is essential for the useful application of large-scale parallel computer systems. As it became clear during the workshop this will gain even more importance in future with the upcoming of peta-scale computing.

 

Mahdi Jalili, Swiss Federal Institute of Technology in Lausanne (EPFL)

Visit Report (University of Melbourne, 9 March 2009 – 2 April 2009)

This visit was made possible through a generous grant by INCF. I visited the Melbourne Neuropsychiatry Center (University of Melbourne, Victoria, Australia) for about 25 days from 9 March 2009 till 2 April 2009. The Melbourne Neuropsychiatry Center headed by Professor Christos Pantelis (http://www.psychiatry.unimelb.edu.au/mnc/about/staff/cpantelis.html) is a joint center among the University of Melbourne, Sunshine Hospital and the Royal Melbourne Hospital. It is a major user of complex computer systems for the storage and analysis of brain images and associated data. Collaboration with other research groups, the Melbourne Neuropsychiatry Center plans to assemble a database of up to 5,000 brain scans and associated clinical data, which could be accessed and examined for morphological markers across disorders including schizophrenia, bipolar disorder, depression, substance abuse disorders, borderline personality disorder, obsessive-compulsive disorder, developmental disorders of adolescence, and other neuropsychiatric conditions. Building and data mining a large-scale database would enable a fundamental shift in how research is conducted into the genesis and development of serious mental illness.

I visited the facilities of the center in Sunshine Hospital in the first couple of days. It was the major place where they were collecting neuroimaging data from schizophrenic (and healthy) subjects. I visited the place where some patients with severe illness had been hospitalized (for two years in some cases). I also discussed with some of the clinical researchers on the methods they collect the psychopathological and neuroimaging data from the subjects. The main format of data collected by the staff of the center was MRI, fMRI and DTI. Researchers in the centre have been examining the 'at-risk' mental state and we have been the first group to demonstrate dynamic brain changes in young people as they develop psychotic illness. I am interested in studying the statistical and dynamical properties of the structural and functional networks of the brain, which has value for “Digital Brain Atlasing Systems” project of INCF. There were some postdocs in the center whose projects were in the same line (Dr. Luca Cocchi and Dr. Andrew Zaleski); I mainly discussed and worked with them for the rest of the visit. They shared their fMRI and DTI data with me and I started exploring the data and looking for possible odd properties of the brain networks extracted from them. I expect to have some preliminary results soon.

The main outcome of the visit was the application for a postdoc grant. I found the center and the research works carried out there interesting and decided to apply for an Australian grant for a postdoctoral training. A grant entitled “Statistical and dynamical properties of brain networks: how psychoses influence brain networks?” was submitted to Australian Health and Medical Research Council that will start from 2010, if granted. We plan to examine brain connectivity in psychotic disorders with the aim of assessing how this changes over time. We will focus particularly on young people at high risk for developing psychosis and in patients with early psychosis. This complements our existing work and will add new directions that will contribute to knowledge about the transition to and evolution of active psychosis, especially schizophrenia.

Summing all, the visit was successful, I believe, and will foster new collaboration for me with a leading group in the field. I would like to thank INCF whose generous support made the visit possible.

Stephen Larson, University of California, San Diego

Purpose:

The major goals for the travel approved for Stephen Larson to travel to the Netherlands, Nov 2009, were:
1. Create a 2D raster image for all brain regions in the Paxinos macaque atlas portion of the scalable brain atlas. The 2D raster should have the appropriate brain region highlighted.
2. Embed this 2D raster image for all Paxinos macaque atlas brain regions in the NeuroLex.
3. Ensure that clicking on the 2D raster image in NeuroLex brings up the appropriate page in the Scalable brain atlas.
4. Programmatically generate entries for primate entities that are named with brain structure and the atlas, e.g., CA1 hippocampal region of PHT00 (Paxinos, Huang, Toga 2000).
5. Incorporate a center point in 3D atlas coordinates for each brain region entry.
6. Move the work on development to production.
In brief; goals 1-4 have been accomplished. Goal 5 has been partially accomplished. Goals 5 & 6 are still in progress and have been substantially advanced by the interactions due to travel.

Detailed description of accomplishments

You can see examples of pages that have been augmented with the results of goals 1-4 on the NeuroLex dev site [1]. These pages feature  the 2D raster image embedded within them for all brain regions defined in the Paxinos macaque atlas. The raster image is a plate from the Paxinos atlas that contains a highlighted region in red that indicates the brain region being described on that page. This highlighting is significant because other systems such as the BrainInfo "where is it?" page, force the user to hunt for the brain region in question via a label, rather than to make it visually clear which region is described. This highlighting functionality also illustrates the beginnings of an ability to transcribe a center point in 3D atlas coordinates for each brain region entry (Goal 5), although the script to perform the transcription is not yet ready.

Goals 2 & 3 required a more careful reconciliation of the Paxinos brain regions and the CoCoMac database than had been done prior to the collaboration. Prior to collaboration, some of the pages did not correctly link, or produce correct images, based on some subtle mismatches for ~30 brain regions. We were able to resolve these mismatches during the collaboration period.

Goal 4 was accomplished in a spreadsheet produced by the collaboration which married the list of Paxinos brain regions provided by the CoCoMac group with new conventions being assembled by the NeuroLex team for a common ontology for mammalian brain anatomy. The spreadsheet is prepared with columns that are appropriate for upload into the NeuroLex. In the next phase of collaboration, the NeuroLex spreadsheet upload tool will be used to put these regions on the NeuroLex.

The final goal, 6, will be accomplished following community review of the work that has been done in development. This is to allow sufficient testing to be done on the system that has been created in order to assure a high degree of satisfaction with the final product. Now that the system is testable, the completion of this goal is well under way.

Additional value to INCF

In addition to the goals that had been pre-approved, several additional value opportunities to the INCF were achieved. The trip co-incided with the Semantic Web Applications and Tools for Life Sciences workshop (http://swat4ls.org), which allowed the NeuroLex to be promoted via a software demo and abstract published in the proceedings. Because of this, the NeuroLex was exposed to ~40 researchers in the Semantic Web Life sciences community. Based on their interest ~10 researchers added themselves to the NeuroLex mailing list. Secondly, the NeuroLex was promoted to the NeuroPI group at the Radboud University. This led to an increased awareness of the activities of the INCF Structural Lexicon task force to this group. The value to the INCF for these opportunities are improved relationships to the Semantic Web community and an improved relationship to this resource for Macaque anatomy and connectivity--a key species for neuroscience research.

1. http://neurolex.incf.net/wiki/Special:Search?search=pht00&go=Find+a+Term