Distribution shifts -- where the training distribution differs from the test distribution -- can substantially degrade the accuracy of machine learning (ML) systems deployed in the wild. Despite their ubiquity, these real-world distribution shifts are under-represented in the datasets widely used in the ML community today. To address this gap, we present WILDS, a curated collection of 8 benchmark datasets that reflect a diverse range of distribution shifts which naturally arise in real-world applications, such as shifts across hospitals for tumor identification; across camera traps for wildlife monitoring; and across time and location in satellite imaging and poverty mapping. On each dataset, we show that standard training results in substantially lower out-of-distribution than in-distribution performance, and that this gap remains even with models trained by existing methods for handling distribution shifts. This underscores the need for new training methods that produce models which are more robust to the types of distribution shifts that arise in practice. To facilitate method development, we provide an open-source package that automates dataset loading, contains default model architectures and hyperparameters, and standardizes evaluations. Code and leaderboards are available at this https URL.

WILDS: A Benchmark of in-the-Wild Distribution Shifts

In order to ensure normal body function, the human body is dependent on a tight control of its blood glucose levels. This is accomplished by a highly sophisticated network of various hormones and neuropeptides released mainly from the brain, pancreas, liver, intestine as well as adipose and muscle tissue. Within this network, the pancreas represents a key player by secreting the blood sugar-lowering hormone insulin and its opponent glucagon. However, disturbances in the interplay of the hormones and peptides involved may lead to metabolic disorders such as type 2 diabetes mellitus (T2DM) whose prevalence, comorbidities and medical costs take on a dramatic scale. Therefore, it is of utmost importance to uncover and understand the mechanisms underlying the various interactions to improve existing anti-diabetic therapies and drugs on the one hand and to develop new therapeutic approaches on the other. This review summarizes the interplay of the pancreas with various other organs and tissues that maintain glucose homeostasis. Furthermore, anti-diabetic drugs and their impact on signaling pathways underlying the network will be discussed.

/pdf/pancreatic-regulation-of-glucose-homeostasis-famtoxtclt.pdf

Pancreatic regulation of glucose homeostasis.

The enteroendocrine system orchestrates how the body responds to the ingestion of foods, employing a diversity of hormones to fine-tune a wide range of physiological responses both within and outside the gut. Recent interest in gut hormones has surged with the realization that they modulate glucose tolerance and food intake through a variety of mechanisms, and such hormones are therefore excellent therapeutic candidates for the treatment of diabetes and obesity. Characterizing the roles and functions of different enteroendocrine cells is an essential step in understanding the physiology, pathophysiology, and therapeutics of the gut-brain-pancreas axis.

Enteroendocrine Cells: Chemosensors in the Intestinal Epithelium.

More than 20 years ago, Pories et al. published a seminal article, "Who Would Have Thought It? An Operation Proves to Be the Most Effective Therapy for Adult-Onset Diabetes Mellitus." This was based on their observation that bariatric surgery rapidly normalized blood glucose levels in obese people with type 2 diabetes mellitus (T2DM), and 10 years later, almost 90% remained diabetes free. Pories et al. suggested that caloric restriction played a key role and that the relative contributions of proximal intestinal nutrient exclusion, rapid distal gut nutrient delivery, and the role of gut hormones required further investigation. These findings of T2DM improvement/remission after bariatric surgery have been widely replicated, together with the observation that bariatric surgery prevents or delays incident T2DM. Over the ensuing two decades, important glucoregulatory roles of the gastrointestinal (GI) tract have been firmly established. However, the physiological and molecular mechanisms underlying the beneficial glycemic effects of bariatric surgery remain incompletely understood. In addition to the mechanisms proposed by Pories et al., changes in bile acid metabolism, GI tract nutrient sensing and glucose utilization, incretins, possible anti-incretin(s), and the intestinal microbiome are implicated. These changes, acting through peripheral and/or central pathways, lead to reduced hepatic glucose production, increased tissue glucose uptake, improved insulin sensitivity, and enhanced β-cell function. A constellation of factors, rather than a single overarching mechanism, likely mediate postoperative glycemic improvement, with the contributing factors varying according to the surgical procedure. Thus, different bariatric/metabolic procedures provide us with experimental tools to probe GI tract physiology. Embracing this approach through the application of detailed phenotyping, genomics, metabolomics, and gut microbiome studies will enhance our understanding of metabolic regulation and help identify novel therapeutic targets.

https://care.diabetesjournals.org/content/diacare/39/6/893.full.pdf

Mechanisms of Diabetes Improvement Following Bariatric/Metabolic Surgery

The gastrointestinal tract has recently come to the forefront of multiple research fields. It is now recognized as a major source of signals modulating food intake, insulin secretion and energy balance. It is also a key player in immunity and, through its interaction with microbiota, can shape our physiology and behavior in complex and sometimes unexpected ways. The insect intestine had remained, by comparison, relatively unexplored until the identification of adult somatic stem cells in the Drosophila intestine over a decade ago. Since then, a growing scientific community has exploited the genetic amenability of this insect organ in powerful and creative ways. By doing so, we have shed light on a broad range of biological questions revolving around stem cells and their niches, interorgan signaling and immunity. Despite their relatively recent discovery, some of the mechanisms active in the intestine of flies have already been shown to be more widely applicable to other gastrointestinal systems, and may therefore become relevant in the context of human pathologies such as gastrointestinal cancers, aging, or obesity. This review summarizes our current knowledge of both the formation and function of the Drosophila melanogaster digestive tract, with a major focus on its main digestive/absorptive portion: the strikingly adaptable adult midgut.

https://www.genetics.org/content/genetics/210/2/357.full.pdf

Anatomy and Physiology of the Digestive Tract of Drosophila melanogaster

Suppression of Insulin Production and Secretion by a Decretin Hormone

Development of accurate disease models and discovery of immune-modulating drugs is challenged by the immune system’s highly interconnected and context-dependent nature. Here we apply deep-learning-driven analysis of cellular morphology to develop a scalable “phenomics” platform and demonstrate its ability to identify dose-dependent, high-dimensional relationships among and between immunomodulators, toxins, pathogens, genetic perturbations, and small and large molecules at scale. High-throughput screening on this platform demonstrates rapid identification and triage of hits for TGF-β- and TNF-α-driven phenotypes. We deploy the platform to develop phenotypic models of active SARS-CoV-2 infection and of COVID-19-associated cytokine storm, surfacing compounds with demonstrated clinical benefit and identifying several new candidates for drug repurposing. The presented library of images, deep learning features, and compound screening data from immune profiling and COVID-19 screens serves as a deep resource for immune biology and cellular-model drug discovery with immediate impact on the COVID-19 pandemic.

/pdf/functional-immune-mapping-with-deep-learning-enabled-5g637gahip.pdf

Kathleen-Rose Skelly

Papers

Suppression of Insulin Production and Secretion by a Decretin Hormone

Functional immune mapping with deep-learning enabled phenomics applied to immunomodulatory and COVID-19 drug discovery