A principal odor map unifies diverse tasks in olfactory perception.

Mapping molecular structure to odor perception is a key challenge in olfaction. We used graph neural networks to generate a principal odor map (POM) that preserves perceptual relationships and enables odor quality prediction for previously uncharacterized odorants. The model was as reliable as a human in describing odor quality: On a prospective validation set of 400 out-of-sample odorants, the model-generated odor profile more closely matched the trained panel mean than did the median panelist. By applying simple, interpretable, theoretically rooted transformations, the POM outperformed chemoinformatic models on several other odor prediction tasks, indicating that the POM successfully encoded a generalized map of structure-odor relationships. This approach broadly enables odor prediction and paves the way toward digitizing odors.

Metabolic activity organizes olfactory representations.

Hearing and vision sensory systems are tuned to the natural statistics of acoustic and electromagnetic energy on earth and are evolved to be sensitive in ethologically relevant ranges. But what are the natural statistics of odors, and how do olfactory systems exploit them? Dissecting an accurate machine learning model (Lee et al., 2022) for human odor perception, we find a computable representation for odor at the molecular level that can predict the odor-evoked receptor, neural, and behavioral responses of nearly all terrestrial organisms studied in olfactory neuroscience. Using this olfactory representation (principal odor map [POM]), we find that odorous compounds with similar POM representations are more likely to co-occur within a substance and be metabolically closely related; metabolic reaction sequences (Caspi et al., 2014) also follow smooth paths in POM despite large jumps in molecular structure. Just as the brain's visual representations have evolved around the natural statistics of light and shapes, the natural statistics of metabolism appear to shape the brain's representation of the olfactory world.

Automated preclinical detection of mechanical pain hypersensitivity and analgesia.

ABSTRACT: The lack of sensitive and robust behavioral assessments of pain in preclinical models has been a major limitation for both pain research and the development of novel analgesics. Here, we demonstrate a novel data acquisition and analysis platform that provides automated, quantitative, and objective measures of naturalistic rodent behavior in an observer-independent and unbiased fashion. The technology records freely behaving mice, in the dark, over extended periods for continuous acquisition of 2 parallel video data streams: (1) near-infrared frustrated total internal reflection for detecting the degree, force, and timing of surface contact and (2) simultaneous ongoing video graphing of whole-body pose. Using machine vision and machine learning, we automatically extract and quantify behavioral features from these data to reveal moment-by-moment changes that capture the internal pain state of rodents in multiple pain models. We show that these voluntary pain-related behaviors are reversible by analgesics and that analgesia can be automatically and objectively differentiated from sedation. Finally, we used this approach to generate a paw luminance ratio measure that is sensitive in capturing dynamic mechanical hypersensitivity over a period and scalable for high-throughput preclinical analgesic efficacy assessment.

Revealing the structure of pharmacobehavioral space through motion sequencing.

Understanding how genes, drugs and neural circuits influence behavior requires the ability to effectively organize information about similarities and differences within complex behavioral datasets. Motion Sequencing (MoSeq) is an ethologically inspired behavioral analysis method that identifies modular components of three-dimensional mouse body language called 'syllables'. Here, we show that MoSeq effectively parses behavioral differences and captures similarities elicited by a panel of neuroactive and psychoactive drugs administered to a cohort of nearly 700 mice. MoSeq identifies syllables that are characteristic of individual drugs, a finding we leverage to reveal specific on- and off-target effects of both established and candidate therapeutics in a mouse model of autism spectrum disorder. These results demonstrate that MoSeq can meaningfully organize large-scale behavioral data, illustrate the power of a fundamentally modular description of behavior and suggest that behavioral syllables represent a new class of druggable target.

Rate of Change of Premeal Glucose Measured by Continuous Glucose Monitoring Predicts Postmeal Glycemic Excursions in Patients With Type 1 Diabetes: Implications for Therapy.

BACKGROUND: Patients with type 1 diabetes routinely utilize a single premeal fingerstick glucose to determine premeal insulin doses. Continuous glucose monitoring (CGM) provides much richer glycemic trend information, including glycemic slope (GS). How to incorporate this information into dosing decisions remains an open question. METHODS: We examined the relationship between premeal GS and postmeal glycemic excursions in 240 individuals with type 1 diabetes receiving CGM augmented insulin pump therapy. Over 23.5 million CGM values were synchronized with 264 500 meals. CGM values were integrated 2 hours premeal to compute GS and 2 hours postmeal to compute glycemic excursion outcomes. Postmeal hyperglycemia (integrated CGM glucose >180 mg/dL*hr) and postmeal hypoglycemic events (any CGM glucose < 70 mg/dL) were tabulated according to positive/negative premeal GS and according to GS bins commonly displayed as rate-of-change arrows on CGM devices. RESULTS: Positive versus negative premeal GS was associated with a 2.28-fold (95% CI 2.25-2.32) risk of postmeal hyperglycemia. Negative versus positive premeal GS was associated with a 2.36-fold (95% CI 2.25-2.43) increase in one or more postprandial hypoglycemic events. Premeal GS in the bin currently displayed as "no change" on existing CGM devices (-1 to 1 mg/dL/min), conferred a 1.82-fold (95% CI 1.79-1.86) risk of postprandial hyperglycemia when positive and a 2.06-fold (95% CI 1.99-2.15) increased risk of postprandial hypoglycemia when negative. CONCLUSION: Premeal GS predicts postmeal glycemic excursions and may help inform insulin dosing decisions. Rate-of-change arrows on existing devices obscure clinically actionable glycemic trend information from CGM users.

Machine Learning Approaches in Cardiovascular Imaging.

Cardiovascular imaging technologies continue to increase in their capacity to capture and store large quantities of data. Modern computational methods, developed in the field of machine learning, offer new approaches to leveraging the growing volume of imaging data available for analyses. Machine learning methods can now address data-related problems ranging from simple analytic queries of existing measurement data to the more complex challenges involved in analyzing raw images. To date, machine learning has been used in 2 broad and highly interconnected areas: automation of tasks that might otherwise be performed by a human and generation of clinically important new knowledge. Most cardiovascular imaging studies have focused on task-oriented problems, but more studies involving algorithms aimed at generating new clinical insights are emerging. Continued expansion in the size and dimensionality of cardiovascular imaging databases is driving strong interest in applying powerful deep learning methods, in particular, to analyze these data. Overall, the most effective approaches will require an investment in the resources needed to appropriately prepare such large data sets for analyses. Notwithstanding current technical and logistical challenges, machine learning and especially deep learning methods have much to offer and will substantially impact the future practice and science of cardiovascular imaging.

Mapping Sub-Second Structure in Mouse Behavior.

Complex animal behaviors are likely built from simpler modules, but their systematic identification in mammals remains a significant challenge. Here we use depth imaging to show that 3D mouse pose dynamics are structured at the sub-second timescale. Computational modeling of these fast dynamics effectively describes mouse behavior as a series of reused and stereotyped modules with defined transition probabilities. We demonstrate this combined 3D imaging and machine learning method can be used to unmask potential strategies employed by the brain to adapt to the environment, to capture both predicted and previously hidden phenotypes caused by genetic or neural manipulations, and to systematically expose the global structure of behavior within an experiment. This work reveals that mouse body language is built from identifiable components and is organized in a predictable fashion; deciphering this language establishes an objective framework for characterizing the influence of environmental cues, genes and neural activity on behavior.

Selective activation of striatal fast-spiking interneurons during choice execution.

Basal ganglia circuits are essential for the organization and execution of voluntary actions. Within the striatum, fast-spiking interneurons (FSIs) are thought to tightly regulate the activity of medium-spiny projection neurons (MSNs) through feed-forward inhibition, yet few studies have investigated the functional contributions of FSIs in behaving animals. We recorded presumed MSNs and FSIs together with motor cortex and globus pallidus (GP) neurons, in rats performing a simple choice task. MSN activity was widely distributed across the task sequence, especially near reward receipt. By contrast, FSIs showed a coordinated pulse of increased activity as chosen actions were initiated, in conjunction with a sharp decrease in GP activity. Both MSNs and FSIs were direction selective, but neighboring MSNs and FSIs showed opposite selectivity. Our findings suggest that individual FSIs participate in local striatal information processing, but more global disinhibition of FSIs by GP is important for initiating chosen actions while suppressing unwanted alternatives.

Opposite effects of stimulant and antipsychotic drugs on striatal fast-spiking interneurons.

Psychomotor stimulants and typical antipsychotic drugs have powerful but opposite effects on mood and behavior, largely through alterations in striatal dopamine signaling. Exactly how these drug actions lead to behavioral change is not well understood, as previous electrophysiological studies have found highly heterogeneous changes in striatal neuron firing. In this study, we examined whether part of this heterogeneity reflects the mixture of distinct cell types present in the striatum, by distinguishing between medium spiny projection neurons (MSNs) and presumed fast-spiking interneurons (FSIs), in freely moving rats. The response of MSNs to both the stimulant amphetamine (0.5 or 2.5 mg/kg) and the antipsychotic eticlopride (0.2 or 1.0 mg/kg) remained highly heterogeneous, with each drug causing both increases and decreases in the firing rate of many MSNs. By contrast, FSIs showed a far more uniform, dose-dependent response to both drugs. All FSIs had decreased firing rate after high eticlopride. After high amphetamine most FSIs increased firing rate, and none decreased. In addition, the activity of the FSI population was positively correlated with locomotor activity, whereas the MSN population showed no consistent response. Our results show a direct relationship between the psychomotor effects of dopaminergic drugs and the firing rate of a specific striatal cell population. Striatal FSIs may have an important role in the behavioral effects of these drugs, and thus may be a valuable target in the development of novel therapies.

Wavelet filtering before spike detection preserves waveform shape and enhances single-unit discrimination.

The isolation of single units in extracellular recordings involves filtering. Removing lower frequencies allows a constant threshold to be applied in order to identify and extract action potential events. However, standard methods such as Butterworth bandpass filtering perform this frequency excision at a cost of grossly distorting waveform shapes. Here, we apply wavelet decomposition and reconstruction as a filter for electrophysiology data and demonstrate its ability to better preserve spike shape. For the majority of cells, this approach also improves spike signal-to-noise ratio (SNR) and increases cluster discrimination. Additionally, the described technique is fast enough to be applied real-time.