Predicting computations necessary for the decoding of odor mixtures by the olfactory system

我们的嗅觉系统处理复杂的气味混合物，来自非常高的 超过10,000种可能气味的维度空间，使用有限的(~100-1000)组 嗅觉受体。虽然已经做了大量的工作来理解 气味是由嗅觉感受器编码的，相反的问题是：“气味是如何 嗅觉系统从感受器的反应中获取气味信息？目前还不清楚。这 该项目旨在确定解码气味所需的计算 来自受体反应的信息。我们将开发解码算法和 感受器气味信息解码的机械神经网络模型 回应。我们将研究这些算法和机制模型的性能 并将他们的预测与嗅觉系统的结构和可用的数据进行比较 关于有机体在嗅觉行为任务中的表现。通过这样的方式 比较，我们将确定解码所需的特定计算 天然气味。这将通过以下目标实现。目标1：发展 用于从受体反应中解码气味信息的算法，并比较它们的 行为数据的性能。目标2：开发生物物理神经网络模型 并将其与气味解码算法进行比较，以预测嗅觉回路的作用。

Our olfactory system processes complex odor mixtures, drawn from a very high dimensional space of over 10,000 possible odorants, using a limited set of (~100-1000) olfactory receptors. While a considerable amount of work has done to understand the odors are encoded by the olfactory receptors, the inverse problem: “How does the olfactory system obtain odor information from receptor response?” is unclear. This project aims to identify the computations that are necessary for the decoding odor information from receptor responses. We will develop decoding algorithms and mechanistic neural network models for decoding odor information from receptor responses. We will study the performance of these algorithms and mechanistic models and compare their predictions to the structure of the olfactory system and available data on the performance of organisms in olfactory behavioral tasks. Through such comparisons, we will identify the specific computations that are necessary for decoding of natural odors. This would be achieved through the following aims. Aim 1: Develop algorithms for decoding odor information from receptor responses and compare their performance to behavioral data. Aim 2: Develop biophysical neural network models the olfactory system and compare it to odor decoding algorithm to predict role of olfactory circuits.