CAREER: From Connectome to Behavior: Computational Models of Multifunctional Neural Circuits in C. elegans

职业：从连接组到行为：线虫多功能神经回路的计算模型

• 批准号: 1845322
• 负责人: Eduardo Izquierdo
• 金额: $ 60万
• 依托单位: Indiana University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-09-01 至 2023-11-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1845322&HistoricalAwards=false
• 关键词: CAREER; Connectome; Behavior; Computational; Models

How an animal flexibly coordinates multiple behaviors as a cohesive unit is one of the central problems of neuroscience; multifunctionality has also been recognized as one of the fundamental challenges in the development of a general artificial intelligence. Although the ability of neural circuits to flexibly reconfigure is widespread among organisms, most studies of the neural basis of behaviors focus on isolated circuits and individual behaviors. Studies that consider multifunctional circuitry tend to focus on the switching between distinct patterns of activity, with little insight into multifunctional sensorimotor integration. With the increasing amount of anatomical, physiological and behavioral data being generated, a computational modeling framework to understand the neural basis of behavior is essential. The goal of this project is to model multiple neural circuits that have been identified in isolation and to integrate them into a single model to better understand how multifunctionality arises in sensory-driven behavioral circuits. This project is an important step toward the long-term goal of developing a behaviorally-functional brain-body-environment model of a living organism at the level of individual neurons. The cross-disciplinary methodologies developed from this project will serve as a springboard for understanding multifunctional circuits in living organisms as well as for generating artificial systems capable of robustly and efficiently performing multiple functions. The project focuses specifically on modeling and analyzing the circuits responsible for the wide range of spatial orientation behaviors in the nematode Caenorhabditis elegans. This model organism is a uniquely qualified target for integrated computational modeling of a complete animal because of the breadth of information known about its genetics, development, anatomy, and behavior. Despite this substantial knowledge, information about the electrophysiological properties of its nervous system is less complete. The project aims to constrain the model by what is known from the anatomy and physiology of the organism with reasoned simplifications about its body and environment. Then, stochastic optimization will be used to fill in electrophysiological unknowns such that the model produces behavior that matches what has been observed, including the effect of neural manipulations on behavior. The result of optimization will not be a unique model, but rather an ensemble of models that are consistent with current knowledge of the system. Each of these possibilities represents a testable hypothesis for C. elegans. The next step in the project will be to analyze the structure of this ensemble to formulate the key experiments that can distinguish between the various classes of possibilities in the worm. The results of such experiments can then be used as additional constraints for subsequent optimizations in an iterative cycle of model refinement. Besides the generation of experimentally-testable predictions that are specific to C. elegans, through the analysis of the ensemble of models, the project aims to discover general principles for how multifunctional circuits operate in living organisms more broadly.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

动物如何灵活地协调多种行为作为一个有凝聚力的单元是神经科学的核心问题之一;多功能性也被认为是通用人工智能发展的基本挑战之一。虽然神经回路灵活重构的能力在生物体中广泛存在，但大多数对行为神经基础的研究都集中在孤立的回路和个体行为上。考虑多功能电路的研究往往集中在不同的活动模式之间的切换，很少深入了解多功能感觉运动整合。随着越来越多的解剖，生理和行为数据的产生，一个计算建模框架，以了解行为的神经基础是必不可少的。该项目的目标是对已被隔离识别的多个神经回路进行建模，并将其整合到一个模型中，以更好地了解多功能性如何在感觉驱动的行为回路中产生。该项目是朝着在单个神经元水平上开发生物体的行为功能性脑-身体-环境模型的长期目标迈出的重要一步。从这个项目开发的跨学科方法将作为一个跳板，了解生物体中的多功能电路，以及生成能够强大和有效地执行多种功能的人工系统。该项目特别侧重于建模和分析负责线虫秀丽隐杆线虫中广泛的空间定向行为的电路。这种模式生物是一个独特的合格的目标，一个完整的动物，因为它的遗传学，发育，解剖学和行为的信息的广度集成计算建模。尽管有这些丰富的知识，关于其神经系统电生理特性的信息还不太完整。该项目旨在通过对生物体的解剖学和生理学的了解来约束模型，并对其身体和环境进行合理的简化。然后，随机优化将用于填充电生理学未知数，使得模型产生与观察到的行为相匹配的行为，包括神经操纵对行为的影响。优化的结果将不是一个独特的模型，而是与系统的当前知识一致的模型的集合。这些可能性中的每一个都代表了C的一个可检验的假设。优雅的该项目的下一步将是分析这个集合的结构，以制定关键实验，可以区分蠕虫中的各种可能性。然后，这些实验的结果可以在模型细化的迭代循环中用作后续优化的附加约束。除了产生特定于C的实验可检验的预测之外。elegans，该项目旨在通过对模型集合的分析，发现多功能电路如何在生物体中更广泛地运作的一般原理。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值进行评估，被认为值得支持和更广泛的影响审查标准。

项目成果

Reinforcement learning beyond the Bellman equation: Exploring critic objectives using evolution

• DOI: 10.1162/isal_a_00338
• 发表时间: 2020-07
• 作者: A. Leite;Madhavun Candadai;E. Izquierdo

Partial copying and the role of diversity in social learning performance

部分复制和多样性在社会学习表现中的作用

• DOI: 10.1177/26339137221081849
• 发表时间: 2022
• 期刊: Collective Intelligence
• 作者: Campbell, Chelsea M;Izquierdo, Eduardo J;Goldstone, Robert L
• 通讯作者: Goldstone, Robert L

Generating reward structures on a parameterized distribution of dynamics tasks

在动态任务的参数化分布上生成奖励结构

• DOI: 10.1162/isal_a_00466
• 发表时间: 2021
• 期刊: Artificial Life Conference Proceedings
• 作者: Leite, Abe;Izquierdo, Eduardo J.
• 通讯作者: Izquierdo, Eduardo J.

What does functional connectivity tell us about the behaviorally-functional connectivity of a multifunctional neural circuit?

关于多功能神经回路的行为功能连接，功能连接告诉我们什么？

• DOI: 10.1162/isal_a_00534
• 发表时间: 2022
• 期刊: ALIFE 2022: The 2022 Conference on Artificial Life
• 作者: Izquierdo, Eduardo J.;Candadai, Madhavun
• 通讯作者: Candadai, Madhavun

Perpetual Crossers without Sensory Delay: Revisiting the Perceptual Crossing Simulation Studies

无感觉延迟的永久交叉者：重新审视感知交叉模拟研究

• DOI: 10.1162/isal_a_00509
• 发表时间: 2022
• 期刊: ALIFE 2022: The 2022 Conference on Artificial Life
• 作者: Izquierdo, Eduardo J.;Severino, Gabriel J.;Merritt, Haily
• 通讯作者: Merritt, Haily