New Computational Neuroscience Textbook

Paul Miller bookComputational Neuroscience is an exciting branch of science, which is helping us understand how simple biophysical processes within cells such as neurons lead to complex and sometimes surprising neural responses, and how these neurons, when connected in circuits can give rise to the wide range of activity patterns underlying human thinking and behavior. To bridge the scales from molecules to mental activity, computer simulations of mathematical models are essential, as it is all too easy for us otherwise to produce descriptions of these complex interacting systems that are internally inconsistent. Simulations allow us to ask “given these ingredients, what is possible?”

Simulation showing how weaker input that is localized can produce spiking when stronger dispersed input does not.

The best way to study computational neuroscience is to write the computer codes that model a particular biological phenomenon, then see what the simulation does when you vary a parameter in the model. Therefore, the course I teach at Brandeis (NBIO 136B) is based around a large number of computer tutorials, in which students, some of whom have no computer-coding background, begin with codes of 5-10 lines that simulate charging of a capacitor, and end up completing codes that simulate the neural underpinnings of learning, pattern recognition, memory, and decision-making. It turns out that very few computational principles are needed to build such codes, making these simulation methods far more easily understood and completed than any mathematical analysis of the systems. However, in the absence of a suitable introductory textbook—most computational neuroscience textbooks are designed by Ph.D. physicists and mathematicians for Ph.D. physicists and mathematicians—it proved difficult for me to use the flipped classroom approach (see below). Therefore, my goal was to create a text that students could read and understand on their own.

Different behaviors of a three-unit circuit as connection-strengths are changed. (Multistable constant activity states, multiple oscillating states, chaotic activity, heteroclinic state sequence). Each color represents firing rate of a unit as a function of time.

In keeping with the goal of the course—to help students gain coding expertise and understand biological systems through manipulations of computer codes—I produced over 100 computer codes (in Matlab) for the book, the vast majority of which are freely available online. (All codes used to produce figures and some tutorial solutions are accessible, but I retained over half of the tutorial solutions in case instructors wish to assign tutorials without students being able to seek a solution elsewhere.)

Learn more at MIT Press.

From the Preface of the book:

I designed this book to help beginning students access the exciting and blossoming field of computational neuroscience and lead them to the point where they can understand, simulate, and analyze the quite complex behaviors of individual neurons and brain circuits. I was motivated to write the book when progressing to the “flipped” or “inverted” classroom approach to teaching, in which much of the time in the classroom is spent assisting students with the computer tutorials while the majority of information-delivery is via students reading the material outside of class. To facilitate this process, I assume less mathematical background of the reader than is required for many similar texts (I confine calculus-based proofs to appendices) and intersperse the text with computer tutorials that can be used in (or outside of) class. Many of the topics are discussed in more depth in the book “Theoretical Neuroscience” by Peter Dayan and Larry Abbott, the book I used to learn theoretical neuroscience and which I recommend for students with a strong mathematical background.

The majority of figures, as well as the tutorials, have associated computer codes available online, at github.com/primon23/Intro-Comp-Neuro, and at my website. I hope these codes may be a useful resource for anyone teaching or wishing to further their understanding of neural systems.

 

Learning to see

How do we learn to see? Proper visual experience during the first weeks and months of life is critical for the proper development of the visual system. But how does experience modify neural circuits so that they exhibit the proper responses to visual stimuli? Knowledge of the mechanisms by which the brain is constructed early in development should inspire new therapies for repairing the brain if it develops improperly or is damaged by disease or injury.

At the present time, it is not possible to directly view all or even most connections within a living neural circuit. Therefore, neuroscientists often build computational models to study how these circuits may be constructed and how they may change with experience. A good model allows scientists to understand how these circuits may work in principle, and offers testable predictions that can be examined in the living animal to either support or refute the model.

Undergraduate Ian Christie ’16 was interested in understanding how neural circuits in the ferret visual system become selective to visual motion. At the time of eye opening, neurons in ferret visual cortex respond to an object moving in either of two opposite directions. With about a week of visual experience, each neuron develops a preference for only one of these directions, and greatly reduces its responses to the opposite direction.

Previous models of this process posited that the primary source of the change was in the organization and pattern of inputs to the cortex. But, recent experiments from the Van Hooser lab (Roy/Osik/Ritter et al., 2016) showed that stimulating the cortex by itself was sufficient to cause the development of motion selectivity, which suggests that some changes within the cortex itself must be underlying the increase in selectivity, at least in part. Further, other experiments in the lab of former Brandeis postdoc Arianna Maffei (Griffen et al., 2012) have shown that the cortex becomes less excitable to focal stimulation over the first weeks after eye opening.

Ian constructed families of computational models that could account for both of these observations. In the model, columns of neurons in the cortex already receive input that is slightly selective for motion in one of two opposite directions, but the connections between these cortical columns are so strong that both columns respond to both directions. However, the activity that is caused by simulated visual experience activates synaptic plasticity mechanisms in the model, that served to greatly reduce the strength of these connections between the columns, allowing motion selectivity to emerge in the cortical columns. The project was supervised by faculty members Paul Miller and Stephen Van Hooser, and the results were published in Journal of Neurophysiology (Christie et al., 2017).

Future experiments will now look for evidence of weaker connectivity between cortical neurons with visual experience.

This work was supported by the “Undergraduate and Graduate Training in Computational Neuroscience” grant to Brandeis University from NIH, and by the National Eye Institute grant EY022122. It also used the Brandeis University High Performance Computing Cluster.

IGERT Video Poster Competition Voting Open

Tony Ng (a grad student in Paul Miller’s lab in Neuroscience) writes:

I’m entering a nationwide video/poster competition organized by the National Science Foundation (NSF) under the IGERT program.  There are over 100 three-minute-videos/posters in the competition.  The videos/posters are divided into 18 fields, all of which are multidisciplinary.  Mine covers cognition/biology/physics.

The competition has a Public Choice award.  Winning the award requires Facebook “likes” on my page.  I need on the order ~1000 “likes” to be in contention.  The bar has been raised from last year’s.  The competition is fierce.  Each/every vote from the Brandeis community counts!

The competition opens today (5/21) and ends Thursday (5/23) at 10pm.  For a vote to count, it is imperative to click on the “Public Choice” button, which would then trigger a Facebook “like” sign-in.  Anyone with an existing Facebook account can contribute.

Here’s the link to my 3-minute video/poster:

http://posterhall.org/igert2013/posters/402

Act now! Tthe competition closes on Thursday at 10pm!

Hope you enjoy the videos!

Update (2 pm):

Andrew Russell from the Petsko-Ringe lab also has a poster in the competition on studying Aβ oligomers to understand Alzheimer’s Disease – check it out — vote early, vote often?

http://posterhall.org/igert2013/posters/416

Dynamic Coding in Neural Signals Workshop on July 29

The Center of Excellence for Learning in Education, Science and Technology (CELEST) is holding a workshop on its cross-function initiative Dynamic Coding in Neural Signals at Boston University (677 Beacon Street, Room B02) on July 29, 2011. from 1:00 – 5:45 pm. The workshop is free and open to the public. There will be talks by invited speakers from 1 – 4:15, including presentations by Don Katz (“Perceptual processing via coherent sequences of ensemble states”) and Paul Miller (“Stochastic transitions between discrete states in models of taste processing and decision-making”). a student and postdoc poster session will follow, with ample opportunity for discussion between presenters and workshop attendees.

CELEST is a joint venture of scientists at four Boston-area universities including Brandeis and is sponspored by the National Science Foundation. Robert Sekuler, Louis and Frances Salvage Professor of Psychology at Brandeis, is a co-Principal Investigator, and Biology and Neuroscience faculty Gina Turrigiano and Paul Miller are also involved in the center.

New in Pubmed

Have no time to write News and Views, but there are a few new papers from our labs that have recently popped up in Pubmed.

Protected by Akismet
Blog with WordPress

Welcome Guest | Login (Brandeis Members Only)