Jonathan W. Pillow, Ph.D. - US grants

This project aims to develop new mathematical and computational tools for understanding the basic information-processing strategies of neurons and neural populations in the brain. Recent technological advances have enabled large-scale recordings of neural activity from intact neural circuits, but there is a severe shortage of theoretical methods for revealing what this activity means--that is, what information it carries, and how it gives rise to behavior. The research described in this proposal will address these questions using novel statistical techniques for studying the neural code in single neurons and neural populations, using both extracellularly and intracellularly recorded neural data.

There are at least two important statistical aspects to the proposed research: first, new methods for reliably estimating the neurobiological variables of interest (e.g., spikes, membrane currents, synaptic conductances, etc.) from noisy experimental recordings; and second, powerful, flexible, model-based methods for understanding the complex, high-dimensional, and time-dependent relationship between sensory stimuli, behavioral responses, and neural activity. The three specific aims of the proposal focus on: (1) the encoding and decoding of decisions from multi-neuron spike trains in parietal cortex; (2) intracellular signals in visual cortex, at the level of membrane potential and synaptic currents, and their relationship to the information conveyed in spike trains; and (3) advanced methods for adaptive, "closed loop" neurophysiology experiments, leading to more informative experimental designs and more interpretable neural datasets. All three aims will involve intensive collaborations with experimental groups and will tightly integrate theory and experiment.

The proposed research will reveal new features of visual and cognitive representations in cortex, and will unlock the neural code at multiple levels of biophysical detail in sensory, motor and cognitive systems. More broadly, the research will shed new light on information flow in groups of neurons, with implications for both the treatment of brain disorders and the design of new technology.

DESCRIPTION (provided by applicant): Intellectual Merit: Perceptual decision-making is an essential cognitive capability. It requires neural circuits that can accumulate sensory evidence, combine it with prior information, and select an appropriate action at an appropriate time. Theories of the brain's ability to perform these computations have primarily involved either mechanistic models based on dynamical systems, or normative models of optimal decision-making imported from psychology, statistics, or economics. However, existing theories do not account-or even attempt to account-for the detailed response properties of neurons believed to carry out these computations. Multi-neuron recordings necessary to evaluate such theories have not yet been collected. There is as yet no general theoretical framework for relating the various sensory, motor, memory, and reward variables involved in decision-making to the time-varying spike responses of multiple neurons that collectively compute decision. This proposal aims to fill that gap. The goal of the proposed research is a detailed and comprehensive understanding of the encoding and decoding of decision-related information by groups of neurons in lateral intraparietal cortex (area LIP), a brain region strongly implicated in decision-making. Multi-electrode recordings will be obtained from primates engaged in decision-making tasks; this will provide the first window into the simultaneous representation of decisions by groups of spiking neurons. The investigators will develop a highly flexible probabilistic spike train model to capture the spike responses of neural populations in LIP, incorporating correlations between neurons, spike-history and adaptation, and a complete set of dependencies on various sensory, motor, decision and reward variables. A novel feature of this research is that it does not presuppose a particular mechanistic or normative theory of LIP function; rather, it begins by seeking a descriptive model of LIP responses as they actually exist in the brain. This will allow for a full accounting of the time-varying information carried by LIP spikes and the optimality of various strategies for decoding them, and will provide a platform for deriving and evaluating simplified models of LIP function. The research will tightly integrate theory and experiment with several new experiments designed to examine the joint coding of decisions across multiple neurons. Collaboration: The proposed research represents a new collaboration between two young investigators with expertise in computational neuroscience and systems neurophysiology. It will combine state-of-the-art statistical methods for spike train modeling and experimental methods recording the simultaneous activity of multiple neurons. The goals of the proposal will be met by closely integrating theory and model development with electrophysiological experiments, which will be facilitated by the proximity of the two investigators. Broader Impacts: The parietal cortex plays a central role in decision-making, and is implicated in a variety of major brain disorders, including depression, anxiety, schizophrenia, and Parkinson's disease. By revealing the computational underpinnings of neural decision making in healthy brains, the proposed research holds great promise for advancing the understanding and treatment of these disorders. Moreover, the models and methodologies to be developed are very general, with applicability to a wide variety of brain areas involved in sensory and motor processing. These methods will aid in the design of advanced sensory and motor neural prosthetic devices, human-engineered systems that replace damaged portions of the sensory or motor system. All software will be made publicly available online, which will enhance the infrastructure for research and education in computational neuroscience. The research proposal will promote teaching and training in several key respects. The project is fundamentally interdisciplinary, combining cutting-edge physiological and computational techniques. Trainees will spend time in both investigator's labs, and will receive an invaluable hands-on, collaborative education. The project will also directly inform classes developed by both investigators. The investigators will promote public scientific understanding by making audio recordings of basic math and science textbooks for the visually impaired at the Learning Ally (Austin's recording studio for the visually impaired). The investigators will aim to recruit interns and graduates from traditionally under-represented groups, especially women. Finally, they will conduct outreach at local middle and high schools in order to spark enthusiasm for mathematics and computer science, disciplines which are fundamental to the exciting challenge of discovering how the brain works.

DESCRIPTION (provided by applicant): Temporal integration may be the brain's key step in generating flexible and intelligent behaviors that are divorced from the immediate time scales of sensory processing or motor execution. This research program aims to unpack the neural computations that underlie higher cognitive function by building off seminal studies in the primate dorsal stream and visual motion processing. Within an integrated framework of psychophysics, neurophysiology, and computation, we aim to ask precise questions about how fleeting sensory signals are read out to guide behavior, and to arrive at answers that span the levels of single neurons, neural circuits, and mental computations. This work strives to understand the basic mechanisms that deteriorate in a variety of mental health and brain aging conditions. Aim 1. Precise characterization of temporal integration from stimulus to decision, using a reverse correlation protocol during psychophysics and simultaneous multiple-neuron / multiple-area recordings across MT and LIP, and interpreted via a generalized linear model. No empirical study has tested the hypothesized temporal integration of MT signals by LIP with direct measurement, which we will do by simultaneously recording from sets of MT and LIP neurons. Aim 2. Causal interrogation of this (putative) decision-making circuit, extending the MT-LIP framework described above to include reversible inactivations of one area, complemented by concurrent multiple-neuron recordings in the other area. The evidence linking LIP to decision-making has been almost entirely correlation. We propose to perform inactivations of the area to test for its causal or necessary role in the accumulation of evidence. Aim 3. Analysis of multiple stages of oculomotor decision-making, applying the multiple-neuron / multiple- area and inactivate-and-record approaches described in the aims above to LIP-FEF-PFC. Although the direct measurement of sensory signals in MT at the same time as LIP recording is likely to grant us significant new insights, the neural signals supporting motion discrimination almost certainly span a larger circuit with many sensory and cognitive stages. Here, we propose to apply the techniques described above to elucidate the relative roles of posterior parietal and prefrontal areas in decision-making.

Project  Summary    Flexible behavior is central to virtually all cognitive and social abilities. Recent technical advances have opened an  unprecedented opportunity to comprehensively dissect the neural circuit mechanisms of this ability across multiple  brain areas in freely behaving animals. This proposal focuses on the cerebellum, a structure that is a major site of  pathology in autism spectrum disorder. Damage to the cerebellum at birth leads to a 36­fold increase in the risk of  autism,  and  this  region  is  also  a  principal  site  for  co­expression  of  autism  risk  genes.  Thus  cerebellar  development  may  act  as  an  intermediate  mechanistic  step  in  transducing  inherited  autism  risk  into  neurodevelopmental  phenotypes.  In  this  project,  a  multidisciplinary  team  of  leading  experts  proposes  to  investigate  the  neural  basis  of  this  disorder  using  advanced  technologies,  including  unbiased  automatic  classification  of  behavior,  large­scale  cellular­resolution imaging in behaving rodents, mouse genetic models for  autism, and manipulation of neural activity in specific cerebellar areas and cell types. In genetic mouse models of  autism,  the  researchers  will  identify  modes  of  behavior  based  on  physical  poses,  and  relate  these  modes  to  classical  behavioral  tests,  such  as  eyeblink  conditioning,  and  to cerebellar circuit dysfunction. In adult wild­type  and autism model mice, the researchers will use optogenetic methods to perturb specific cerebellar lobules while  quantifying  the  effects  on  behavioral  dynamics  and  learning.  In  juvenile  model  mice,  the  researchers  will  use  chemogenetic methods to identify long­lasting patterns of behavioral disruption and relate these patterns across  behaviors to build a quantitative map of these perturbations. In addition, they will use in vivo dendritic imaging to  evaluate the influences of cerebellar perturbation on neocortical neuron structure. All of these results will inform  modeling  of  cerebellar­neocortical  interactions  to  better  understand how these differently wired regions interact  during learning and development. The long­term goal of this project is to arrive at a chain of explanation, centered  on  principles  of  convergent  neuroscience,  to  understand  causal mechanisms of neurodevelopmental disorders.  This  project  will  join  genetics  with  circuit  function,  local  cerebellar  anatomy  with  behavioral  outcomes,  and  classical  behavioral  tests  with  modern  unbiased  methods.  This project is expected to produce an accurate and  detailed  understanding  of  cerebellar  contributions  to  normal  and  aberrant  neurodevelopment.  In  addition,  the  proposed research will enable researchers to generate and test a variety of hypotheses about the neural basis of  flexible  behavior.  Taken  together,  these  achievements  will  represent  a  crucial  step  toward  a  mechanistic  understanding  of  how the brain develops its complex ability to respond flexibly to the environment, from birth to  adulthood.   

Project Summary Working memory, the ability to temporarily hold multiple pieces of information in mind for manipulation, is central to virtually all cognitive abilities. Recent technical advances have opened an unprecedented opportunity to comprehensively dissect the neural circuit mechanisms of this ability across multiple brain areas. The task to be studied is a common form of decision-making that is based on the gradual accumulation of sensory evidence and thus relies on working memory. A team of leading experts propose to investigate the neural basis of this behavior using the latest techniques, including virtual reality, high-throughput automated behavioral training, large-scale cellular-resolution imaging in behaving rodents, manipulation of neural activity in specific brain areas and cell types, and automated anatomical reconstruction. In particular, the researchers will identify key brain regions that are required for this decision task through systematic, temporally specific inactivations via optogenetics technology, across all of dorsal cortex and in key subcortical areas, and use quantitative model-fitting to evaluate the effects. They will use state-of-the-art two-photon calcium imaging methods and electrophysiology to characterize the information flow in many individual neurons within these brain areas during the task. In addition, they will use cutting-edge anatomical reconstructions and new functional connectivity methods, within and across brain regions, to evaluate the interactions of these physiologically characterized neurons. The long-term goal of this project is to arrive at a complete, brain-wide understanding of the cellular and circuit mechanisms of activity dynamics related to working memory. Finally, they will use sophisticated computational methods to incorporate this new understanding into a realistic circuit model that will support a tightly integrated process of model-guided experimental design, in which the model suggests the most informative experiments and their results are then fed back to improve the model?s fidelity. This process is expected to produce the most accurate and detailed multi-brain-region biophysical circuit model of a cognitive process in existence. In addition, the proposed research will enable researchers to generate and test a variety of hypotheses about the neural basis of evidence accumulation, working memory, and decision-making. Taken together, these achievements will represent a crucial step toward a mechanistic understanding of how the brain works with information.

Project Summary: Project 5, Analysis and Modeling of Neural Data Working memory, the ability to temporarily hold multiple pieces of information in mind for manipulation, is central to virtually all cognitive abilities. This multi-component research project aims to comprehensively dissect the neural circuit mechanisms of this ability across multiple brain areas. Large population recordings, such as those that will be obtained in other components of this proposal, open the door to assessing the dynamics of brain states on a single-trial, moment-by-moment basis. Yet their size and complexity present a challenge, as does the variety of data that will be collected, incorporating anatomy, behavior, neural activity, and perturbations. This project will develop and apply novel statistical analyses and modeling approaches to meet these challenges. The lion?s share of the variance in neural population activity is often dominated by variations in a small number of variables, which are called ?latent variables.? This project will leverage the very large data sets, collected in other components of the project, of many simultaneously recorded neurons to develop advanced linear and nonlinear methods to identify the most informative latent variables. To analyze these datasets, the researchers will develop new latent variable discovery methods. First, they will combine advanced quantitative behavioral analysis with advanced statistical neural analysis. Second, they will combine latent space discovery with fitting of generalized linear models to neural data. The resulting nonlinear methods will provide an unprecedentedly complete statistical description of the data: these methods aim to simultaneously discover and capture the dynamics of the most important latent variables, and to produce a full statistical characterization of the responses of each individual recorded neuron. In biophysical modeling work, critical to creating a mechanistic understanding at the neural circuit level, this project will develop and test models of both local and multi-brain-region activity during working memory and decision-making. These models will build upon rigorous sensitivity-analysis techniques for identifying the critical network interactions underlying observed behavior. The models will be used both to interpret existing data and to design maximally informative experiments about inter-regional network interactions, and they will provide a principled platform from which to design future experiments that test specific hypotheses about function and further constrain the models.

Project Summary/Abstract Social interactions across the animal kingdom, from courtship rituals and aggressive interactions to spoken conversation, are wondrously complex - they necessarily involve back- and-forth feedback between nervous systems transmitted through multiple sensory modalities and each animal's behavior. Typical experiments in this field observe only a tiny fraction of the activity in any neuronal circuit, and then only under a very limited range of behavioral conditions. To overcome this limitation, the proposed research leverages the compact nervous system of Drosophila melanogaster, combined with its wealth of genetic tools, to study the dynamic behavioral interactions and detailed neural mechanisms that underlie courtship between males and females. The project combines unbiased measurement of behavior, neural circuit manipulations, neural recordings in behaving animals, and sophisticated computational models. The specific aims include: i) elucidating the computations that the brain performs during courtship by mapping the sensorimotor transformations underlying male and female interactions over time via quantitative behavioral assays and the generation of predictive models; ii) combining models with neural perturbations to map the underlying circuits that govern the link between sensory inputs and behaviors; And, iii) testing the models of neural control during courtship by monitoring neural activity in behaving animals experiencing fictive courtship stimuli in a virtual-reality apparatus. This work will substantially advance our understanding of how two interacting brains process and transfer information, and will uncover general principles of neural circuit function that will inform studies of sensorimotor integration in more complex animals, such as rodents and humans. The project will also produce new experimental and theoretical tools for studying social behaviors. Finally, it will shed light on the mechanisms that go awry in several disorders, including Autism Spectrum Disorder (ASD), in which sensory perception becomes disentangled from motor outputs ? these disorders have profound effects on cognitive well-being and a major impact on public health.

Project Summary / Abstract High-throughput experimental neuroscience has made it possible to observe behavior of many animals, as well as a large groups of neurons simultaneously, providing an exciting opportunity for figuring out how the neural system performs computations that underlie perception, cognition, and behavior. However, there is a major bottleneck in the scientific cycle of data analysis and data collection due to the complexity and scale of noisy, high-dimensional data. The primary objective of this project is to develop tools for tracking the internal state of the brain that are not directly measurable from both the behavior and neural signals, and to generate optimal stimulus corresponding to the current brain state. These external stimuli can be used to perturb the animal?s belief or strategy about the world such that the animal would behave differently. Aim 1: Our team will develop a neural state tracking system that will parse out and display complex neural signals recorded from the animal brain in real-time. The neural state tracking algorithm will also extract the law that the neural system operates under, allowing neuroscientist to generate a new class of hypotheses about the population level implementation underlying intelligent behavior. Aim 2: To causally test hypothesis on how population of neurons compute and produce meaningful behavior, it is necessary to be able to perturb the internal computation process. We will develop a feedback control system to perturb the neural dynamics at a short time scale with a novel control scheme for neural computation that respects the brain?s own degrees of freedom. Aim 3: By understanding and tracking the time evolution of internal strategy throughout learning, we can learn how to optimize the training of animal behavior. In this aim, we will develop statistical models of learning and a computational system to generate the best stimuli based on the past performance of the animal. The statistical tools developed in this project will likely accelerate fundamental discoveries in neuroscience. Clinically, this research can extend to monitoring, diagnosing, and building next- generation real-time feedback stimulation devices for disorders with a neurodynamic or behavioral component such as Parkinson?s disease, autism, learning disorders, obsessive compulsive disorder, and epilepsy.

Summary/Abstract, Core D: Behavioral Analysis and Modeling This proposal?s overarching goal is to understand how internal states influence decisions and to identify the underlying neural mechanisms. The Behavioral Analysis and Modeling Core?s development, testing, and application of statistical tools to rigorously characterize behavioral states is critical to achieving this goal. This collaboration will study behavioral state changes defined on three different time scales: those arising spontaneously with engagement and disengagement in a task, those resulting from changing expectations during the task, and those resulting from learning within and across days. The goals of this core are to develop and extend novel open-source analytical tools for extracting state information from behavioral and video data over these three timescales. First, the investigators will identify latent states governing choice behavior, which vary across trials within an experimental session, using tools based on a hidden Markov model with generalized linear model outputs. In addition, they will develop a hierarchical extension of the model to take statistical advantage of the vast behavioral dataset produced by the proposed experimental projects. Next, they will infer behavioral states that vary within a single trial using cutting-edge video analysis methods. In particular, they will apply state-of-the-art markerless tracking methods to extract the position of animal features (paws, tongue, nose, etc.) from behavioral video, and extend these methods to obtain estimates of animal pose in three dimensions (fusing multiple camera views). They will then combine the markerless tracking output with nonlinear autoencoder compression methods to obtain a more informative semi-supervised, low-dimensional data representation of the video data. Using machine learning methods, they will temporally segment the resulting representation to obtain interpretable behavioral states within each trial (e.g., ?rest,? ?groom,? ?reach?), suitable for further downstream analyses. Finally, they will develop new tools to track the dynamics of behavior over the course of learning. Decision-making strategies evolve during training, both within and across sessions, and continue to vary even in well-trained animals. To characterize these state changes, the investigators will develop and apply novel statistical models that combine state-space modeling and reinforcement learning approaches to analyze the learning curves observed in individual animals as they are trained to perform the International Brain Laboratory decision-making task. The resulting framework will quantify how much of the pronounced differences in learning curves across animals can be attributed to differences in identified learning rules, and will help identify neural correlates of inferred learning dynamics in brainwide recordings. All software tools that are developed will be fully open source and will be shared via a public, parallelized cloud implementation for maximal scalability and reproducibility.