Keynote Lectures:


Excitatory and inhibitory population activity that guides perceptual decisions

Ann Churchland

Decisions are driven by the coordinated activity of excitatory and inhibitory 
populations in multiple neural structures. Inhibitory neurons play a critical 
role in many models of decision-making, but the difficulty in measuring large 
inhibitory populations in behaving animals has left their in vivo role mysterious. 
To understand the contributions of excitatory and inhibitory neural populations to 
perceptual decision-making, we measured neural responses in the posterior parietal 
cortex (PPC) of transgenic mice expressing tdTomato in inhibitory neurons (GAD2-Cre 
crossed with Ai14 reporter line). To record neural activity, mice were injected with 
AAV9-Synapsin-GCaMP6f in the posterior parietal cortex. Mice were trained to make 
perceptual decisions about multisensory stimuli. Specifically, head-fixed mice were 
presented with a series of multisensory “events” (clicks and flashes). They were 
trained to lick to a right (left) port for event rates above (below) an abstract 
category boundary (16 Hz). We then used 2-photon imaging to measure single-neuron 
responses during these decisions. In each session, ~600 neurons were simultaneously 
recorded while mice performed ~400 trials. After correcting for bleed through from the 
green (pan-neuronal) channel to the red (inhibitory) channel, inhibitory neurons were 
identified using a combination of local contrast and correlation measures. Having identified 
which measured neurons were excitatory vs. inhibitory, we used a linear classifier to 
determine the information carried by each population about different features of the decision. 
We observed that on single trials, both excitatory and inhibitory populations could 
reliably predict the animal's current as well as previous choice. Interestingly, the 
inhibitory population carried as much information as the excitatory population about the 
animal's choice. Also the classifier weights for the inhibitory population were similarly 
heterogeneous as the excitatory population. These findings open the venue to distinguish 
candidate decision-making models that predict specific vs non-specific roles for how 
inhibitory neurons modulate the activity of excitatory neurons.


Cellular modalities controlling basal ganglia output

Alon Korngreen

 The Leslie and Susan Gonda Interdisciplinary Brain Research Center, 
 and the Mina and Everard Goodman Faculty of Life Sciences 
 Bar Ilan University, Ramat Gan 52900, Israel

The entopeduncular nucleus is one of the basal ganglia output nuclei 
integrating synaptic information from several pathways within the basal ganglia. 
The firing of EP neurons is modulated by two streams of inhibitory synaptic 
transmission, the direct pathway from the striatum and the indirect pathway 
from the globus pallidus. These two inhibitory pathways continuously modulate 
the firing of EP neurons. However, the link between these synaptic inputs to 
neuronal firing in the EP is unclear.  To investigate this input-output 
transformation we performed whole-cell and perforated-patch recordings from 
single neurons in the entopeduncular nucleus in rat brain slices during 
repetitive stimulation of the striatum and the globus pallidus at frequencies 
within the in vivo activity range of these neurons. These recordings, 
supplemented by compartmental modeling, showed that GABAergic synapses 
from the striatum, converging on EP dendrites, display short-term 
facilitation and that somatic or proximal GABAergic synapses from the 
globus pallidus show short-term depression. Activation of striatal synapses 
during low presynaptic activity decreased postsynaptic firing rate by 
continuously increasing the inter-spike interval. Conversely, activation of 
pallidal synapses significantly affected postsynaptic firing during high 
presynaptic activity. Our data thus suggest that low frequency striatal output 
may be encoded as progressive phase shifts in downstream nuclei of the basal 
ganglia while high frequency pallidal output may continuously modulate EP firing. 
Next, we used Immunohistochemistry and qRT-PCR to determine the types and 
distribution of dopamine receptors in the EP. These experiments demonstrated that 
all DR subtypes were highly expressed by EP neurons suggesting that, in the EP, 
DRs act postsynaptyicaly. The anatomical data suggested that both D1-like receptors 
(D1LRs) and D2-like receptors (D2LRs) would affect information processing in the 
EP. Indeed, whole-cell recordings revealed that striatal inputs to the EP were 
potentiated by D1LRs whereas  pallidal inputs to the EP were depressed by D2LRs. 
We further show that these changes in synaptic efficacy changed the information 
content of EP neuron firing. Thus, our findings suggest that the dopaminergic system 
affects the passage of feedforward information through the BG by modulating 
input divergence in the striatum and output convergence in the EP.


Extracting the interplay of excitation and inhibition, from
intracellular recordings and populations of single units in cerebral

Alain Destexhe, UNIC, CNRS, Gif sur Yvette, France

We show two ways of extracting information about excitatory and
inhibitory activities, and their interplay, in cerbral cortex.  First,
from intracellular recordings, it is possible to extract the
distributions of excitatory and inhibitory inputs from measuring the
membrane potential (Vm).  The Vm recording can be treated as due to
the interplay of two stochastic processes, excitation and inhibition,
and the main parameters of these processes can be extracted.  It is
also possible in some cases to extract the full time course of
excitatory and inhibitory inputs from the Vm activity.  We show that
applying these methods to intracellular recordings in awake animals
reveals that most spikes are caused by a dis-inhibition rather than
due to an increase of excitation.  Second, excitatory and inhibitory
activities can also be extracted from unit recordings obtained with
multi-electrode arrays.  Regular-spiking (RS) and fast-spiking (FS)
units can be discriminated, and their excitatory or inhibitory nature
determined from functional interactions.  We show that RS and FS cells
display different patterns of correlations, and they are tightly
balanced in all brain states, except during epileptic seizures, during
which the two populations dissociate.  FS cells are also more rhythmic
than RS cells, and they are strongly correlated with fast oscillations
in the beta and gamma range.  Finally, FS cell activity seems to be
the main determinant of local field potentials in cerebral cortex.  We
conclude that at the level of single neurons and populations of
neurons, inhibitory neurons are much more than simple followers of
excitation, they are heavily involved in all properties investigated.

Human seizures couple across spatial scales through traveling wave dynamics

Mark A. Kramer

Epilepsy - the propensity toward recurrent, unprovoked seizure - is a devastating 
disease affecting 65 million people worldwide. Understanding and treating this 
disease remains a challenge, as seizures manifest through mechanisms and features 
that span spatial and temporal scales. In this talk, we will examine some aspects 
of this challenge through the analysis and modeling of human brain voltage activity 
recorded simultaneously across microscopic and macroscopic spatial scales. We will 
show some evidence that rapidly propagating waves of activity sweep across the cortex 
during seizure. We will also describe a corresponding computational model to propose 
specific mechanisms that support the observed spatiotemporal dynamics.

Andrew Leifer

Probing neural drivers of behavior in the nematode C. elegans

How does a simple nervous system generate animal behavior?  We are pursuing this 
question using a suite of optical neurophysiology tools we have developed to manipulate 
and measure neural activity in the brain of the simple nematode C. elegans as it crawls 
freely. I will describe two investigations: one studying sensorimotor transformations 
in the worm and the other investigating whole brain neural activity during spontaneous 
and unrestrained behavior. Both investigations highlight the challenges inherent in 
interpreting large scale recordings of neural activity and behavior and demonstrate the 
need for statistical inference and machine learning approaches. 


Treating epilepsy using team science and human intracranial electrophysiology

Brian Litt

 Human intracranial electrophysiology provides a path to understanding brain networks that generate seizures, 
 and to developing new therapies. Because of the heterogeneity of individual cases and their electrode placements, 
 understanding the problem and potential solutions requires integrating and sharing data across many patients, 
 different institutions, platforms and centers. This is only viable when nomenclature, imaging, electrode localization, 
 annotation and data collection methods and analysis code are well documented, of high quality and shared openly. This 
 is difficult in scientific environments where incentives to share and accomplish large goals together are misaligned. 
 In this talk I will detail the challenges associated with team science using human intracranial electrophysiology, 
 give examples of powerful rewards that can be obtained when these issues are conquered, and challenge the group to 
 work together on a Common Data Ecosystem for Human Electrophysiology to rapidly advance our fields. While I will draw 
 on my experience working in human epilepsy for examples, I will link these issues to work in cognitive neuroscience 
 and related areas. 

Analysis and modeling of olfactory navigation behavior in walking Drosophila.

Katherine Nagel

Many organisms use airborne odor plumes to locate food or mates.  Plume tracking 
is challenging because turbulence causes the plume to fluctuate in space and time.  
Most animals therefore use a combination of wind and odor cues to navigate towards an 
attractive odor in natural settings.  We seek to understand the computations necessary 
to turn dynamic wind and odor stimuli inputs into a successful, goal-directed behavior.  
To this end, we have developed a system of miniature windtunnels that allow us to precise 
control the wind and odor stimuli experienced by freely-walking fruitflies.  Using this 
paradigm, we find that flies show three types of behavior.  In the absence of odor, flies 
exhibit search behavior with a downwind bias.  In the presence of an attractive odor, they 
walk straighter and faster upwind.  Following loss of odor, they exhibit local search 
behavior, characterized by increased turning. To clarify the contributions of wind- and 
odor-sensing to these behaviors, we stabilized the antennae, which prevents flies from 
sensing wind currents.  In these flies, we found that the baseline downwind bias was lost, 
as well as upwind orientation during odor.  However, flies still walked straighter and 
faster during odor, and searched when odor was lost, implying both pure olfactory, as well 
as multi-sensory contributions to navigation behavior.  Based on these data, we develop a 
simple computational model of olfactory navigation and explore the ability of this model to 
locate odor in diverse environments.  Ongoing studies aim to identify the neural circuit 
mechanisms underlying 
these computations.

Reverse engineering neocortical intelligence

Andreas Tolias
Rice University

Over several decades, the quest to advance artificial intelligence (AI) has used different approaches such as 
symbolic reasoning, expert systems, statistics and neural networks.  Recently, deep learning stirred a renaissance 
in AI by reaching human — or even superhuman — performance on several tasks. From the media and the gaming 
industry to the internet, mobile devices, autonomous machines, security and defense, deep learning is transforming 
industries at an accelerating rate. Deep learning networks have notable similarities to the brain, involving many 
layers, many neurons, and many plastic synapses that change with experience. Yet they differ significantly in 
important respects, lacking cell types, complex nonlinearities, pervasive feedback, structured connectivity, and 
local learning rules. The fact that the most successful artificial networks share such important features with the 
brain, yet have so many differences in their details, suggests that there are enormous opportunities to revolutionize 
machine learning and build next-generation AI systems by understanding and incorporating features derived from 
neuroscience into artificial neural networks. I will describe our ongoing experimental and computational efforts and 
statistical challenges we face to decipher the algorithms of cortical microcircuits and how we are beginning to transfer 
these algorithms to advance machine learning.     

Think Global, Connectome Local: The Analytical Power of the Local Connectome

Timothy D. Verstynen Ph.D., 

Like all intelligent systems, the human brain is constrained by the details of its design. 
In this talk, we will show how the local architecture of white matter pathways within a voxel 
and between adjacent voxels, called the local connectome, provides a data structure for 
characterizing white matter pathways. we will show how it is possible to characterize the 
uniqueness of local connectomic systems in an individual, as well as show how this 
uniqueness is driven by both genetic an environmental factors. we will go on to show how 
it is possible to characterize similarities between individuals along social, health, and 
cognitive dimensions through specific patterns in the local connectome. Finally we show 
preliminary work on the impact that proximal variation in white matter pathways has on 
global activity flow in macroscopic brain networks.

Joint work with Fang-Cheng Yeh M.D., Ph.D.