Anne K. Churchland - US grants

DESCRIPTION (provided by applicant): The goal of this proposal is to extend our understanding of decision-making beyond the simple paradigms that have been studied thus far. Physiological recording and behavioral analysis will be used to describe how monkeys respond to more complex decision tasks;theoretical modeling will help interpret the responses. My training thus far makes me ideally suited for this endeavor: I have extensive experience in conducting physiology experiments in awake, behaving monkeys. Further, I have some experience in theoretical modeling;additional training will prepare me to make advances as an independent investigator. As a postdoctoral fellow, I have collected physiological data from monkeys engaged in a multiple choice decision task. This data has the potential to make a major advance to our understanding of decision making: the simplicity of previous tasks used to study decision-making suggests that they may offer limited insight into decision-making in general. Indeed, my data pose a major challenge to current models. The proposed career development plan incorporates training from my current mentor, Dr. Shadlen, on theoretical techniques to extend current models to explain more complex decisions. My co-mentors, Drs. Wang and Pouget, have agreed to enter into a collaboration to address this question. The collaboration will allow me to consider two novel models. The combined training from three mentors will provide a foundation in theoretical neuroscience that I can combine with my experimental skills as an independent researcher. For my independent research, I propose to combine my experimental and theoretical skills on a new series of experiments that further test more complex decisions. These experiments combine the visual decision-making task from my postdoctoral work with a novel, auditory decision-making task. The goals of this work are twofold: first, I will ask how the models of decision-making can be extended to incorporate input from 2 modalities. Next, I will ask how behavioral and physiological responses change when decisions are based on evidence from two modalities. A number of clinical disorders, foremost among them Autism, appear to cause impairments in the ability to integrate sensory information. Identifying neural mechanisms that underlie sensory integration within and across modalities may inform clinical treatments for Autistic patients.

The ultimate goal of this project is to provide a novel model of the cognitive and neural basis of numerical cognition, and to use this knowledge to guide the development of new training methods that could improve mathematical abilities in children. The project is a collaboration among investigators at the University of Rochester, Johns Hopkins University, and Cold Spring Harbor Laboratories. Recent research suggests that acuity of numerosity judgments is predictive of success in formal mathematics education, and that similar cognitive processes can be trained by specific kinds of domain-general experience. The core idea is that the firing of neurons encodes a probability distribution, thereby representing simultaneously the most probable sample from the distribution and the variance (i.e., confidence) of the estimate.

This project will develop and test a formal Bayesian model that has the unique feature of naturally accounting for a number of metacognitive factors, a critical but undertested factor in the acquisition of expertise. The primary advantages of this Bayesian approach are its ability to provide a natural description of: 1) how the confidence of a learner relates to the precision of their number knowledge; 2) how a learner can combine information from multiple sources of information about number; 3) how intuitive preferences (also known as prior belief) predict learners' errors; and 4) how improvements in probabilistic inference may benefit the precision of the number sense.

A Research Experience for Undergraduates (REU) Sites award has been made to Cold Spring Harbor Laboratory (CSHL) that will provide research training for 8 students, for 10 weeks during the summers of 2012- 2014. The program trains participants on the present and growing need to integrate biological research with sophisticated computational tools and techniques. CSHL has over 40 faculty members, including members of a newly established Quantitative Biology Department, who will serve as bioinformatics and computational biology mentors in fields ranging from plant biology to machine learning for biology. Through this NSF-REU support, students are afforded the opportunity to conduct full-time research in an appropriately matched lab based on mutual interests and goals. CSHL REU participants have access to individual and shared laboratory facilities such as flow cytometry, high throughput sequencing and analysis, imaging, and proteomics facilities. Participants attend multiple seminars and workshops, such as the responsible conduct in research, professional communication skills, the graduate school application process, and introduction to science careers. REU participants also are invited to attend the CSHL summer courses or meetings, which cover a range of topics such as Computational Neuroscience and Single Cell Analysis. All students are housed on campus within walking distance of their laboratories and the CSHL cafeteria, where they receive the majority of their meals. The multilayer recruitment effort consists of both traditional and digital mailings to potential students and their professors, as well as recruitment visits to universities throughout the country. Students are selected based on academic record, motivation for the proposed program of study, and potential as future researchers. Alumni successes are monitored to determine their continued interest in their academic field of study, their career paths, and the long-term impact of their research experience. Information about the program will be assessed using faculty and student evaluations, as well as the use of an REU common assessment tool. More information is available by visiting http://www.cshl.edu/education/urp/nsf-sponsored-reu-in-bioinformatics-and-computational-biology, or by contacting the PI (Dr. Zachary Lippman at lippman@cshl.edu) or the co-PI (Dr. Doreen Ware at ware@cshl.edu).

DESCRIPTION (provided by applicant): Decisions about visual stimuli are frequently shaped by inputs from other sensory modalities. The goal of this proposal is to gain a deeper understanding of the neural mechanisms that enable integration of visual and auditory inputs for decision-making. Behavioral experiments have established that subjects can integrate information across sensory modalities to make better decisions. Further, subjects weight each sensory modality in proportion to its reliability. Experimenters can estimate subjects' perceptual weights by presenting conflicting information to two sensory modalities and examining the degree to which decisions are biased toward one modality or the other. Despite a wealth of data about the behavioral consequences of multisensory stimuli, much remains unknown about the underlying neural mechanisms. By collecting electrophysiological and behavioral data together, we are in an ideal position to connect multisensory decision-making to its underlying neural mechanism. My central hypothesis is that neurons will show greater stimulus-driven modulation for multisensory stimuli than for unisensory stimuli and that the neural weights that we estimate by comparing unisensory and multisensory responses will be similar for reliable and unreliable stimuli. The posterior parietal cortex is a candidate area for supporting multisensory improvements in rats: it receives inputs from auditory/visual areas and plays a role in motor planning. In Aim 1, we will estimate humans' and rats' perceptual weights on a novel decision-making task in which subjects judge the overall rate of a stream of events: i.e. flashes, clicks or both together. In Aim 2, we will collect electrophysiological data from posterior parietal corte of rats while they are engaged in the task to establish that responses reflect decision-related activity. In Aim 3, we will measure responses for unisensory vs. multisensory stimuli at different levels of stimulus reliability. For each level of reliability, we will estimate the neural weights that describe the degree of stimulus-driven modulation for auditory vs. visual inputs. If the neural weights, like the perceptual weights, change with reliability, this would suggest that stimulus reliability is explicitly encoded by single neurons. If, instead, the neural weights remain unchanged for high- vs. low-reliability stimuli, this would suggest that reliability is automaticaly encoded by a population response that naturally become more variable when it has a lower gain. The latter possibility is predicted by an optimal model of decision-making.

This REU Site award to Cold Spring Harbor Laboratory (CSHL), located in Cold Spring Harbor, NY, will support the training of ten students for ten weeks during the summers of 2016-2018. This award is supported by the Division of Biological Infrastructure in the Directorate for Biological Sciences (BIO) and the Division for Mathematical Sciences in the Directorate for Mathematics and Physical Sciences (MPS).CSHL's REU in Bioinformatics and Computational Neuroscience (BCN) provides participants with an exceptional research experience, integrating genomics and neuroscience through shared analysis tools. Spanning genomes, cells, organisms and the brain, the program trains students to approach complex biological systems quantitatively. Students conduct full-time, independent research under the mentorship of one of CSHL's approximately 50 faculty members working in genomics, quantitative biology, and neuroscience. Participants have access to state-of-the-art technologies, such as high-throughput sequencing and two-photon imaging, and attend lab meetings and research seminars. The REU curriculum includes workshops on quantitative techniques, responsible conduct of research, scientific communication, and scientific careers. The REU culminates with a symposium in which participants present their work to CSHL's scientific community. Students are housed on CSHL's 110-acre campus, within walking distance of laboratories and dining halls. Participants receive room and board and a summer stipend and have access to campus amenities. Students apply online, supplying a personal statement, two letters of recommendation, and academic records. REU participants are selected based on academics, motivation, and demonstrated potential.

It is anticipated that 30 students, primarily from schools with limited research opportunities or those from underrepresented groups, will be trained in CSHL's REU in BCN. Participants will learn to interrogate biological questions with computational tools and techniques. Through the 10-week REU experience, participants will learn how research is conducted. Many will present the results of their work at national scientific conferences, furthering their identity as independent scientists.

A common web-based assessment tool used by all REU programs funded by the Division of Biological Infrastructure (Directorate for Biological Sciences) is used to determine the effectiveness of the training program. Students are tracked after the program to determine career paths. Students will be asked to respond to an automatic email sent via the NSF reporting system. More information about the program is available by visiting http://www.cshl.edu/Education/NSF-REU-in-Bioinformatics-and-Computational-Neuroscience.html, or by contacting the PI (Dr. Anne Churchland, churchland@cshl.edu), the co-PI (Dr. Michael Schatz, mschatz@cshl.edu).

The goal!of the proposed research is to uncover the brain wide circuits and local computations that together allow animals to combine multiple, diverse sources of information to guide decision-making. Specifically, these experiments will investigate how mammals integrate sensory signals over time and across sensory modalities. The main hypothesis is that in addition to sense-specific circuits, auditory and visual decisions rely on common, core decision circuits for decision-related computations, such as evidence accumulation and action selection. Within these structures, targeted connectivity between excitatory and inhibitory neurons supports persistent activity and competition for action selection. The proposed method to test this hypothesis is to measure and manipulate neural activity in mice trained to make perceptual decisions about auditory and visual stimuli. Three approaches, taken together, form the core of the proposal to evaluate this hypothesis and provide a new view of decision-making circuits. First, the degree to which auditory and visual decisions activate overlapping or largely separate neural structures will be evaluated based on wide field imaging of cortex-wide activity during decision-making. Cortex-wide activity will be measured in transgenic mice that express calcium indicators in cortical excitatory neurons. Classifiers and decision-making models we will used to link activity in a given brain structure to decision-making computations. This approach will uncover candidate areas that are active during auditory, visual or multisensory decisions. Next, optogenetic suppression of these candidate areas will be used to evaluate their causal role in decision-making. A model-based comparison of behavior on suppression and control trials will evaluate the effects of disruption on decision-related computations such as event discrimination, evidence accumulation, and action planning. Finally, areas that are identified as causal for specific decision computations will be investigated more closely to understand how these computations are implemented by single neurons. 2- photon microscopy will be used to image populations of single neurons. The experimental subjects will be transgenic mice in which inhibitory neurons emit red fluorescent light that is independent of the green fluorescence that is used as an estimate of neural activity. These two separate signals make it possible to distinguish excitatory from inhibitory neurons and evaluate their respective roles in decision-making. Single-trial classifiers will be used to evaluate the ability of excitatory and inhibitory populations to predict the animal?s choice. The outcome of this experiment will be used to distinguish candidate models of decision-making.