Search for: All records

Abstract Most real-world behaviors – such as odor-guided locomotion - are performed with incomplete information. Activity in olfactory receptor neuron (ORN) classes provides information about odor identity but not the location of its source. In this study, we investigate the sensorimotor transformation that relates ORN activation to locomotion changes inDrosophilaby optogenetically activating different combinations of ORN classes and measuring the resulting changes in locomotion. Three features describe this sensorimotor transformation: First, locomotion depends on both the instantaneous firing frequency (f) and its change (df); the two together serve as a short-term memory that allows the fly to adapt its motor program to sensory context automatically. Second, the mapping between (f, df) and locomotor parameters such as speed or curvature is distinct for each pattern of activated ORNs. Finally, the sensorimotor mapping changes with time after odor exposure, allowing information integration over a longer timescale. 

ABSTRACT Odors released from mates and resources such as a host and food are often the first sensory signals that an animal can detect. Changes in locomotion in response to odors are an important mechanism by which animals access resources important to their survival. Odor-modulated changes in locomotion in insects constitute a whole suite of flexible behaviors that allow insects to close in on these resources from long distances and perform local searches to locate and subsequently assess them. Here, we review changes in odor-mediated locomotion across many insect species. We emphasize that changes in locomotion induced by odors are diverse. In particular, the olfactory stimulus is sporadic at long distances and becomes more continuous at short distances. This distance-dependent change in temporal profile produces a corresponding change in an insect's locomotory strategy. We also discuss the neural circuits underlying odor modulation of locomotion. 

Changes in locomotion mediated by odors (odor-guided locomotion) are an important mechanism by which animals discover resources important to their survival. Odor-guided locomotion, like most other behaviors, is highly variable. Variability in behavior can arise at many nodes along the circuit that performs sensorimotor transformation. We review these sources of variability in the context of the Drosophila olfactory system. While these sources of variability are important, using a model for locomotion, we show that another important contributor to behavioral variability is the stochastic nature of decision-making during locomotion as well as the persistence of these decisions: Flies choose the speed and curvature stochastically from a distribution and locomote with the same speed and curvature for extended periods. This stochasticity in locomotion will result in variability in behavior even if there is no noise in sensorimotor transformation. Overall, the noise in sensorimotor transformation is amplified by mechanisms of locomotion making odor-guided locomotion in flies highly variable. 

We present research in the modeling of neurons within Drosophila (fruit fly) olfaction. We describe the process from data collection, to model creation, and spike generation. Our approach utilizes computational elements such as spiking neural networks that employ leaky integrate-and-fire neurons with adaptive firing behavior that more closely mimick biological neurons. We describe the methods of several learning implementations in both software and hardware. Finally, we present both quantitative and qualitative results on learning spiking neural network models. 

Many behaviors that we perform everyday, including something as familiar as making a peanut-butter sandwich, consist of a sequence of recognizable acts. These acts may include, for example, holding a knife and opening a jar. Yet often neither the sequence nor the individual acts are always performed in the exact same way. For example, there are many ways to hold a knife and there are many ways to open a jar, meaning neither of these actions could be called “stereotyped”. A lack of stereotypy makes it difficult for a computer to automatically recognize the individual acts in a sequence. This same problem would apply to other common behaviors, such as walking around somewhere you have not visited before. While we easily recognize it when we see it, walking is not a stereotyped behavior. It consists of a series of movements that differ between individuals, and even in the same individual at different times. So how can someone automatically recognize the individual acts in a non-stereotyped behavior like walking? To begin to find out, Tao et al. developed a mathematical model that can recognize the walking behavior of a fruit fly. Existing recordings of fruit flies walking were analyzed using a type of mathematical model called a Hierarchical Hidden Markov Model (often shortened to HHMM). Such models assume that there are hidden states that influence the behaviors we can see. For example, someone’s chances of going skiing (an observable behavior) depend on whether or not it is winter (a hidden state). The HHMM revealed that the seemingly random wanderings of a fly consist of ten types of movement. These include the “meander”, the “stop-and-walk”, as well as right turns and left turns. Exposing the flies to a pleasant odor – in this case, apple cider vinegar – altered how the flies walked by changing the time they spent performing each of the different types of movement. All flies in the dataset used the same ten movements, but in different proportions. This means that each fly showed an individual pattern of movement. In fact, the differences between flies are so great that Tao et al. argue that there is no such thing as an average walk for a fruit fly. The model represents a complete description of how fruit flies walk. It thus provides clues to the processes that transform an animal’s sensory experiences into behavior. But it also has potential clinical applications. Similar models for human behaviors could help reveal behaviors that are abnormal because of disease. Normal behaviors also show variability, and some diseases increase or decrease this variability. By making it easier to detect these changes, mathematical models could support earlier diagnosis of medical conditions.