Travel grant reports 2013
Jochen Martin Eppler, Jülich Research Centre, Germany
During my stay from 3rd through 5th Juli 2013 at the INCF Secretariat in Stockholm, Mikael Djurfeldt and I were working on topics related to the interface for the Connection Set Algebra (CSA) in the neuronal simulation tool NEST. In detail, we were planning the outline for our joint manuscript on the connection generator interface for the Research Topic "Python in Neuroscience II" in Front. Neuroinform., which is meanwhile submitted and under review. In order to have a clean separation between the simulator and the connectivity-generating library (e.g. CSA), we created a general library (libneurosim) on the INCF Software Center that can be used and extended by the community. During the visit in Stockholm, we wrote the first code for libneurosim and moved all relevant parts from both NEST's and CSA's code bases to that library. The move necessitated a refactoring of the code of both sides, which we were partly carrying out on site. The visit was concluded by planning the next steps in the collaboration.
Thanks to the support from INCF, I attended the CSHL Asia Summer School in Beijing in July 2013. At the course, I attended tutorials on a variety of computational modeling problems. These included sessions on single neuron models (Hodgkin-Huxley, Fitzhugh-Nagumo, Izhikevich, etc), the lobster pyloric rhythm, divisive normalization models of attention, recurrent dynamics in working memory models, analysis of dynamical systems, and reinforcement learning. We also had a number of high quality lectures by leading computational and experimental neuroscientists covering new research using a variety of computational methods. In the second half of the course, I completed an independent modeling project with the guidance of XJ Wang from NYU. For this project, I built and evaluated a network (rate)
model of prefrontal cortex function during a paired associates task using the framework developed by Rigotti and colleagues (Rigotti et al., 2010). Specifically, I evaluated this model
in terms of its ability to account for observed similarity structure of population single-unit responses in macaque prefrontal cortex in this task as measured by Sigala & colleagues
(Sigala et al., 2008). My analysis of the model highlighted the assumptions one needs to make about the way the network represents the task, for it to be able to account for the data. Prior to this course, I had limited and largely informal experience in computational modeling. On the other hand, I have a varied experimental background in neuroscience,
with undergraduate training and some research experience at the molecular/cellular level, as well as extensive graduate training and experience at the systems/cognitive level. Attending this course and getting hands-on experience with modeling has helped me fill a key gap in my training and is already, significantly, influencing my research program.
Purpose: attend the INCF Congress in Stockholm and to visit collaborators at the Nencki Institute in Warsaw
Travel report: In 2013, I was awarded a travel grant by the INCF to attend the INCF Congress in Stockholm and to visit collaborators at the Nencki Institute in Warsaw. The cost of flights from Australia is normally quite prohibitive to the attending international conferences and indeed the cost of flights slightly exceeded the travel grant of $2500US. However, contributions from my institution meant I had sufficient funding for conference registration and accommodation. I decided to make the most of the trip and also attend the Introduction to Neuroinformatics course that preceded the main congress. The lectures were highly informative and covered a the wide range of topics that make up Neuroinformatics. Previously, I had been unsure of the scope of the term, as different people use it in different ways. For example, I had previously attended the Neuroinformatics course at Woods Hole, MA, USA, and whilst it was a very rewarding course it focussed on specific issues related to analysis of time series data in neuroscience. The congress itself was excellent and had impressive array of speakers. The first workshop, “Analysis and interpretation of massively parallel electrophysical data” was particularly useful for me. The work of Jonathan Pillow in reverse correlation methods and the work of Matteo Carandini in visual neuroscience are very relevant to the work in our lab. I was very happy to be able to ask a scientist of Professor Carandini’s calibre a question at the end of his talk. There were many other interesting and relevant talks, in particular those on digital brain atlases and data sharing and storage. I presented a poster on digital brain atlasing which was great opportunity to demonstrate the lab’s work as well as meeting others involved in similar projects, such as the rat Waxholm project. Finally, the special session on “Large scale brain initiatives” was probably the highlight for me, especially the presentation from Clay Reid on the new “MindScope” project at the Allen Institute. For me, this one of most exciting areas in neuroscience and where I would like to take my career. After the congress, I visited the Neuroinformatics lab of Daniel Wojtek at the Nencki Institute in Warsaw. The main purpose of this project was to discuss a collaborative project we had been working on, in which we digitised a marmoset monkey cortical atlas. After the visit, the data was finally made public on their website, 3dbar.org. We discussed how to proceed the next stage of project, to digitize the subcortical structures for the public release. In addition, we also found we have similar interests in the reconstruction of 3D brain volumes from histology. The work of Piotr Majka in the oppossum was extremely impressive and will assist us in similar projects with our marmoset data. As a result of this, Piotr is now planning to visit our lab in Australia to discuss his methodology and to develop further collaborations. In summary, the travel grant provided a unique and special opportunity for me to attend a world class conference, and further expand our existing international collaborations.
Purpose: To visit and collaboration with Prof. Shun-ichi Amari for working on a project with title “Spatial phase structure of natural images”
The main aim to visit RIKEN was to continue our previous research on the statistical model of the phase information of natural scenes in the visual cortex. We also worked on the temporal phase information of neural oscillators. In the first research, our focus was on the principle that visual sensory system is adapted efficiently to the regularities of natural scenes. We explore the circular structure in the phase spectrum of natural images. We report the presence of structured bimodal distributions for the phase variables of the complex receptive fields responding to natural scenes, suggesting that the uniform distribution is insufficient to reduce the redundancy of natural scenes. Subsequently, we tried to use such statistical phase structures as prior to capture higher-order causes of the redundancy in natural images. Moreover, we follow our study in defining the complex blind source separation. The results of this project are accepted in the annual meeting of the vision science society (VSS14). I there gave a talk on our previous results and ongoing activities in the computational vision. We received useful comments from audience for future works. In the second project, we provide a method to study the phase causality in the brain based on the neural measurements. We propose a new probabilistic method to estimation directed parameters of the kuramoto model from noisy multivariate circular time series in the reverse problem framework. The results are presented in the annual Computational and Cognitive Neuroscience meeting (Cosyne 14). During my stay at RIKEN, I had weekly meetings with Prof. Amari about our projects. He always gave me many insights into our projects.
In addition, I participated in the scientific activities and seminars on the Brain sciences and learned about some projects at RKEN in computational neuroscience and analysis of neural data. I also had a good discussion with RIKEN members on topics which directly relevant to my research projects. In summary, I familiarized with the RIKEN neuroscience community, unique scientific environment, which will be really an excellent way to increase my knowledge in future. Moreover, the internship program was a very rich scientific experience for me to deepen my understanding in mathematical neuroscience by seeing how researchers are tackled problems using alternative approaches. I would like to express my utmost gratitude to the INCF and RIKEN Brain Science Institute for supporting my participation.
Purpose: To visit Dr. Alan Evans’ Laboratory at the MNI, Montreal and learn and implement analysis of structural MRI data using the CIVET pipeline.
Activities and Accomplishments: As part of this two-week visit, I learnt how to analyse structural MRI data to measure cortical thickness and surface area using the automated CIVET pipeline (http://www.bic.mni.mcgill.ca/ServicesSoftware/CIVET), developed at the McConnell Brain Imaging Centre, MNI. I worked with a post-doctoral fellow, Budhachandra Khundrakpam, in Dr.Evans’ lab. The primary goal was to learn and implement this analysis on a dataset of children with ASD (Autism Spectrum Disorders) and neurotypical controls, which I had acquired in India as part of my Ph.D. The broader objective was to be able to implement this analysis back in India, using the pipeline on the online CBRAIN portal (cbrain.mcgill.ca), to be able to conduct cross-cultural studies using common measures. I spent the first few days learning how to analyse the data, about the various data formats and quality control checks that needed to be implemented, as well as the pros and cons of the analysis. The next week was spent analysing my own data and interpreting the results in contest of our previous functional imaging findings. I also got a chance to interact with many of the members of MNI and was able to discuss my research in the context of neuroimaging in autism. On the last day of my visit, I presented some of the results I had generated during my visit to Montreal, as well as some of my earlier work on auditory processing in Autism. We also discussed the possibility of a future collaboration to continue implementation of this analysis once I was back in India.
Overall, the visit was a great learning experience and gave me chance to learn a new method of neuroimaging analysis which I can now share with my colleagues in India. Additionally, I also got to interact with some of the leading scientists in the fields and got hands-on training in one of the state-of-the-art methods in neuroimaging research. I am extremely grateful to INCF for having given me this opportunity to visit Dr.Evans’ lab and get this experience. It will be an invaluable addition to my graduate training.
Mikael Lundqvist - Department of Computational Biology, Stockholm University and KTH Royal Institute of Technology, Sweden
Purpose: to initiate a collaboration between our department and the Carmena lab at the Department of Electrical Engineering and Computer Sciences, University of California, Berkeley. In particular, my objective was to gain insights and enhance practical understanding of experimental work undertaken in the Carmena lab to improve our modelling methodology.
Activities and accomplishments: I visited Carmena's lab for 10 days, from the until the of July, 2013. During this time I interacted mostly with Dr Ryan Canolty and Preeya Khanna, who was directly responsible for the acquisition electrophysiological recordings using multi-unit electrode arrays from monkeys' motor cortex in a brain-machine interface paradigm. Our discussions were aimed at finding best ways to analyze the data and use them to constrain the cortical model development at my home department, where I mostly collaborate with Dr Pawel Herman and Prof. Anders Lansner. The stay at the University of California also facilitated interaction with other experimental and theoretical groups. I presented the current state-of-the-art in neural modelling at KTH to Prof. Robert Knight's group. Due to their interest in our work and future plans made with Prof Carmena's lab we managed to create a broader framework for collaborating on large-scale mutli-scale modelling of the cortex. In this regard we have had fruitful discussions about hypotheses and predictions that can be verified in the combined experimental and modelling studies. We agreed on the type of electrophysiological recordings that should be made so that we could maximise the relevance of the data for modelling purposes. The attractor network modelling approach I proposed was also accepted with enthusiasm. Besides the foundations for a fruitful collaboration and planned methodological advancements in integrating biological data with computational work, we committed ourselves to aim for a publication study in a high impact journal. Finally, I also believe that my visit to the University of California, Berkeley led to the increased recognition of our research group at Stockholm University and KTH Royal Institute of Technology as well as that of the co-founder of this initiative - INCF.
Relevance to INCF: The project undertaken in collaboration with Dr Jose Carmena's group at the Department of Electrical Engineering and Computer Sciences at the University of California, Berkeley bears relevance to the INCF multi-scale modeling program. One of the key objectives is to develop a large-scale model that spans and accounts for different levels of biological data - from cellular and extracellular recordings to behaviour. Besides the scientific value of the project, which lies in the enhanced understanding of neural correlates of motor cortex dynamics and function, the expected outcome also embraces methodological issues. Within the project it is intended to develop a large-scale modeling approach with integrated analysis and synthesis of electrophysiological data, which should facilitate more direct comparisons with biological and synthetic data. This bridge between different sources of data has been initiated during my visit.
Purpose: To attend the 2013 Berkeley summer course in mining and modeling of neuroscience data at UC-Berkeley’s Redwood center for computational neuroscience. In attending this course, I sought to improve the breadth and depth of my understanding of neuroscience while improving my programming and computational modeling skills.
Activities: The course brought together around 30 students with various backgrounds, including computer science, neuroscience, pure and applied math and electrical engineering, with the aim to gain collaborative skills and approaches for tackling problems in neuroscience using quantitative tools. The course consisted of daily lectures delivered by six different lecturers, Robert Kass, Johnathan Pillow, Odelia Schwartz, Sonja Grun, Maneesh Sahani and Tom Dean, combined with daily hands-on problem solving in the form of computational exercises using MATLAB. The focus of the lecturers ranged from problems of ‘encoding’ and ‘decoding’, to modeling spike chain correlation across multiple arrays and trials, to information theory, statistical methodology and more. Each daily lecture and exercise was followed by an evening presentation by a bay area neuroscience researcher. The evening lectures covered topics from spontaneous retinal waves, optogenetics, olfactory processing, neural network modeling, brain-machine interfacing and more. Videos of lectures will become available here: http://redwood.berkeley.edu/videos.php
Results: I found the richness and diversity of the participant’s background fostered great discussions about everything from experimental design, signal processing theory, and memory consolidation to more nuts and bolts aspects of neuroscience such as spike sorting. Furthermore, the diversity of participant focuses within the field of neuroscience from patch clamp experimenters, to encephalopathy researchers, to computer modelers of simulated data made it clear to me how many directions are needed to gain understanding the cross-disciplinary field of neuroscience. Due to the great diversity in backgrounds, I learned a large cross-section of what was possible in computational neuroscience that expands what I can accomplish in my own lab. Through meeting a post-doc who measures coherence between low frequency field potentials and spiking activities, I’m able to start work on both gathering and relating low and high frequency potentials on my neural cultures. Secondly, by interacting with multiple researchers experienced in batch processing data for spike sorting, I’ve successfully sorted spikes on my own data thereby improving the quality of the data. Thirdly, by meeting another researcher focusing on across trial and network spike correlations, I can apply such techniques to my own data to better detect changes induced by chemicals. Thank you INCF for the travel grant!