No abstract

Background: Alzheimer's disease (AD) is the most common type of dementia, and patients with advanced AD frequently lose the ability to identify family members. The fusiform gyrus (FUS) of the brain is critical in facial recognition. However, AD etiology in the FUS of AD patients is poorly understood. New analytical strategies are needed to reveal the genetic and epigenetic basis of AD in FUS. Results: A complex of new analytical paradigms that integrates an array of transcriptomes and methylomes of normal controls, AD patients, and "AD-in-dish" models were used to identify genetic and epigenetic signatures of AD in FUS. Here we identified changes in gene expression that are specific to the FUS in brains of AD patients. These changes are closely linked to key genes in the AD network. Profiling of the methylome (5mC/5hmC/5fC/ 5caC) at base resolution identified 5 signature genes (COL2A1, CAPN3, COL14A1, STAT5A, SPOCK3) that exhibit perturbed expression, specifically in the FUS and display altered DNA methylome profiles that are common across AD-associated brain regions. Moreover, we demonstrate proof-of-principle that AD-associated methylome changes in these genes effectively predict the disease prognosis with enhanced sensitivity compared to presently used clinical criteria. Conclusions: This study identified a set of previously unexplored FUS-specific AD genes and their epigenetic characteristics, which may provide new insights into the molecular pathology of AD, attributing the genetic and epigenetic basis of FUS to AD development.

A longstanding goal of the field of AI is a strategy for compiling diverse experience into a highly capable, generalist agent. In the subfields of vision and language, this was largely achieved by scaling up transformer-based models and training them on large, diverse datasets. Motivated by this progress, we investigate whether the same strategy can be used to produce generalist reinforcement learning agents. Specifically, we show that a single transformer-based model -with a single set of weights -trained purely offline can play a suite of up to 46 Atari games simultaneously at close-to-human performance. When trained and evaluated appropriately, we find that the same trends observed in language and vision hold, including scaling of performance with model size and rapid adaptation to new games via fine-tuning. We compare several approaches in this multi-game setting, such as online and offline RL methods and behavioral cloning, and find that our Multi-Game Decision Transformer models offer the best scalability and performance. We release the pre-trained models and code to encourage further research in this direction. 1

The COVID‐19 pandemic caused by SARS‐CoV‐2 virus quickly spread globally, infecting over half a billion individuals, and killing over 6 million*. One of the more unusual symptoms was patients' complaints of sudden loss of smell and/or taste, a symptom that has become more apparent as the virus mutated into different variants. Anosmia and ageusia, the loss of smell and taste, respectively, seem to be transient for some individuals, but for others persists even after recovery from the infection. Causes for COVID‐19‐associated chemosensory loss have undergone several hypotheses. These include non‐functional or destroyed olfactory neurons and gustatory receptors or of their supporting cells, disruption of the signaling protein Neuropilin‐1, and disruption in the interaction with semaphorins, key molecules in the gustatory and olfactory axon guidance. The current paper will review these hypotheses and chart out potential therapeutic avenues.

Prompted models have demonstrated impressive few-shot learning abilities. Repeated interactions at test-time with a single model, or the composition of multiple models together, further expands capabilities. These compositions are probabilistic models, and may be expressed in the language of graphical models with random variables whose values are complex data types such as strings. Cases with control flow and dynamic structure require techniques from probabilistic programming, which allow implementing disparate model structures and inference strategies in a unified language. We formalize several existing techniques from this perspective, including scratchpads / chain of thought, verifiers, STaR, selection-inference, and tool use. We refer to the resulting programs as language model cascades.

A major pathophysiological cause of cardiovascular disease is vascular plaque calcification. Fluorine 18–Sodium Fluoride (18F-NaF) PET/CT can be used as a sensitive imaging modality for detection of vascular calcification. The aim of this study was to find a non-invasive, cost-efficient, and readily available metric for predicting vascular calcification severity. This retrospective study was performed on 36 participants who underwent 18F-NaF fused PET/CT scans. The mean standard uptake values (SUVs) were calculated from manually sectioned axial sections over the aortic arch and thoracic aorta. Correlation analyses were performed between SUVs and calculated atherogenic indices (AIs). Castelli’s Risk Index I (r = 0.63, p < 0.0001), Castelli’s Risk Index II (r = 0.64, p < 0.0001), Atherogenic Coefficient (r = 0.63, p < 0.0001), Atherogenic Index of Plasma (r = 0.51, p = 0.00152), and standalone high-density lipoprotein (HDL) cholesterol (r = −0.53, p = 0.000786) were associated with aortic calcification. AIs show strong association with aortic arch and thoracic aorta calcifications. AIs are better predictors of vascular calcification compared to standalone lipid metrics, with the exception of HDL cholesterol. Clinical application of AIs provides a holistic metric beneficial for enhancing screening and treatment protocols.

No abstract

No abstract