Foundation models for generalist medical artificial intelligence.

The exceptionally rapid development of highly flexible, reusable artificial intelligence (AI) models is likely to usher in newfound capabilities in medicine. We propose a new paradigm for medical AI, which we refer to as generalist medical AI (GMAI). GMAI models will be capable of carrying out a diverse set of tasks using very little or no task-specific labelled data. Built through self-supervision on large, diverse datasets, GMAI will flexibly interpret different combinations of medical modalities, including data from imaging, electronic health records, laboratory results, genomics, graphs or medical text. Models will in turn produce expressive outputs such as free-text explanations, spoken recommendations or image annotations that demonstrate advanced medical reasoning abilities. Here we identify a set of high-impact potential applications for GMAI and lay out specific technical capabilities and training datasets necessary to enable them. We expect that GMAI-enabled applications will challenge current strategies for regulating and validating AI devices for medicine and will shift practices associated with the collection of large medical datasets.

Human mobility networks reveal increased segregation in large cities.

A long-standing expectation is that large, dense and cosmopolitan areas support socioeconomic mixing and exposure among diverse individuals1-6. Assessing this hypothesis has been difficult because previous measures of socioeconomic mixing have relied on static residential housing data rather than real-life exposures among people at work, in places of leisure and in home neighbourhoods7,8. Here we develop a measure of exposure segregation that captures the socioeconomic diversity of these everyday encounters. Using mobile phone mobility data to represent 1.6 billion real-world exposures among 9.6 million people in the United States, we measure exposure segregation across 382 metropolitan statistical areas (MSAs) and 2,829 counties. We find that exposure segregation is 67% higher in the ten largest MSAs than in small MSAs with fewer than 100,000 residents. This means that, contrary to expectations, residents of large cosmopolitan areas have less exposure to a socioeconomically diverse range of individuals. Second, we find that the increased socioeconomic segregation in large cities arises because they offer a greater choice of differentiated spaces targeted to specific socioeconomic groups. Third, we find that this segregation-increasing effect is countered when a city's hubs (such as shopping centres) are positioned to bridge diverse neighbourhoods and therefore attract people of all socioeconomic statuses. Our findings challenge a long-standing conjecture in human geography and highlight how urban design can both prevent and facilitate encounters among diverse individuals.

Scientific discovery in the age of artificial intelligence.

Artificial intelligence (AI) is being increasingly integrated into scientific discovery to augment and accelerate research, helping scientists to generate hypotheses, design experiments, collect and interpret large datasets, and gain insights that might not have been possible using traditional scientific methods alone. Here we examine breakthroughs over the past decade that include self-supervised learning, which allows models to be trained on vast amounts of unlabelled data, and geometric deep learning, which leverages knowledge about the structure of scientific data to enhance model accuracy and efficiency. Generative AI methods can create designs, such as small-molecule drugs and proteins, by analysing diverse data modalities, including images and sequences. We discuss how these methods can help scientists throughout the scientific process and the central issues that remain despite such advances. Both developers and users of AI toolsneed a better understanding of when such approaches need improvement, and challenges posed by poor data quality and stewardship remain. These issues cut across scientific disciplines and require developing foundational algorithmic approaches that can contribute to scientific understanding or acquire it autonomously, making them critical areas of focus for AI innovation.

Organization of the human intestine at single-cell resolution.

The intestine is a complex organ that promotes digestion, extracts nutrients, participates in immune surveillance, maintains critical symbiotic relationships with microbiota and affects overall health1. The intesting has a length of over nine metres, along which there are differences in structure and function2. The localization of individual cell types, cell type development trajectories and detailed cell transcriptional programs probably drive these differences in function. Here, to better understand these differences, we evaluated the organization of single cells using multiplexed imaging and single-nucleus RNA and open chromatin assays across eight different intestinal sites from nine donors. Through systematic analyses, we find cell compositions that differ substantially across regions of the intestine and demonstrate the complexity of epithelial subtypes, and find that the same cell types are organized into distinct neighbourhoods and communities, highlighting distinct immunological niches that are present in the intestine. We also map gene regulatory differences in these cells that are suggestive of a regulatory differentiation cascade, and associate intestinal disease heritability with specific cell types. These results describe the complexity of the cell composition, regulation and organization for this organ, and serve as an important reference map for understanding human biology and disease.

Toward universal cell embeddings: integrating single-cell RNA-seq datasets across species with SATURN.

Analysis of single-cell datasets generated from diverse organisms offers unprecedented opportunities to unravel fundamental evolutionary processes of conservation and diversification of cell types. However, interspecies genomic differences limit the joint analysis of cross-species datasets to homologous genes. Here we present SATURN, a deep learning method for learning universal cell embeddings that encodes genes' biological properties using protein language models. By coupling protein embeddings from language models with RNA expression, SATURN integrates datasets profiled from different species regardless of their genomic similarity. SATURN can detect functionally related genes coexpressed across species, redefining differential expression for cross-species analysis. Applying SATURN to three species whole-organism atlases and frog and zebrafish embryogenesis datasets, we show that SATURN can effectively transfer annotations across species, even when they are evolutionarily remote. We also demonstrate that SATURN can be used to find potentially divergent gene functions between glaucoma-associated genes in humans and four other species.

Mobility network models of COVID-19 explain inequities and inform reopening.

The coronavirus disease 2019 (COVID-19) pandemic markedly changed human mobility patterns, necessitating epidemiological models that can capture the effects of these changes in mobility on the spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)1. Here we introduce a metapopulation susceptible-exposed-infectious-removed (SEIR) model that integrates fine-grained, dynamic mobility networks to simulate the spread of SARS-CoV-2 in ten of the largest US metropolitan areas. Our mobility networks are derived from mobile phone data and map the hourly movements of 98 million people from neighbourhoods (or census block groups) to points of interest such as restaurants and religious establishments, connecting 56,945 census block groups to 552,758 points of interest with 5.4 billion hourly edges. We show that by integrating these networks, a relatively simple SEIR model can accurately fit the real case trajectory, despite substantial changes in the behaviour of the population over time. Our model predicts that a small minority of 'superspreader' points of interest account for a large majority of the infections, and that restricting the maximum occupancy at each point of interest is more effective than uniformly reducing mobility. Our model also correctly predicts higher infection rates among disadvantaged racial and socioeconomic groups2-8 solely as the result of differences in mobility: we find that disadvantaged groups have not been able to reduce their mobility as sharply, and that the points of interest that they visit are more crowded and are therefore associated with higher risk. By capturing who is infected at which locations, our model supports detailed analyses that can inform more-effective and equitable policy responses to COVID-19.

Annotation of spatially resolved single-cell data with STELLAR.

Accurate cell-type annotation from spatially resolved single cells is crucial to understand functional spatial biology that is the basis of tissue organization. However, current computational methods for annotating spatially resolved single-cell data are typically based on techniques established for dissociated single-cell technologies and thus do not take spatial organization into account. Here we present STELLAR, a geometric deep learning method for cell-type discovery and identification in spatially resolved single-cell datasets. STELLAR automatically assigns cells to cell types present in the annotated reference dataset and discovers novel cell types and cell states. STELLAR transfers annotations across different dissection regions, different tissues and different donors, and learns cell representations that capture higher-order tissue structures. We successfully applied STELLAR to CODEX multiplexed fluorescent microscopy data and multiplexed RNA imaging datasets. Within the Human BioMolecular Atlas Program, STELLAR has annotated 2.6 million spatially resolved single cells with dramatic time savings.

Postmortem memory of public figures in news and social media.

Deceased public figures are often said to live on in collective memory. We quantify this phenomenon by tracking mentions of 2,362 public figures in English-language online news and social media (Twitter) 1 y before and after death. We measure the sharp spike and rapid decay of attention following death and model collective memory as a composition of communicative and cultural memory. Clustering reveals four patterns of postmortem memory, and regression analysis shows that boosts in media attention are largest for premortem popular anglophones who died a young, unnatural death; that long-term boosts are smallest for leaders and largest for artists; and that, while both the news and Twitter are triggered by young and unnatural deaths, the news additionally curates collective memory when old persons or leaders die. Overall, we illuminate the age-old question of who is remembered by society, and the distinct roles of news and social media in collective memory formation.

MARS: discovering novel cell types across heterogeneous single-cell experiments.

Although tremendous effort has been put into cell-type annotation, identification of previously uncharacterized cell types in heterogeneous single-cell RNA-seq data remains a challenge. Here we present MARS, a meta-learning approach for identifying and annotating known as well as new cell types. MARS overcomes the heterogeneity of cell types by transferring latent cell representations across multiple datasets. MARS uses deep learning to learn a cell embedding function as well as a set of landmarks in the cell embedding space. The method has a unique ability to discover cell types that have never been seen before and annotate experiments that are as yet unannotated. We apply MARS to a large mouse cell atlas and show its ability to accurately identify cell types, even when it has never seen them before. Further, MARS automatically generates interpretable names for new cell types by probabilistically defining a cell type in the embedding space.

Graph-based clinical recommender: Predicting specialists procedure orders using graph representation learning.

OBJECTIVE: To determine whether graph neural network based models of electronic health records can predict specialty consultation care needs for endocrinology and hematology more accurately than the standard of care checklists and other conventional medical recommendation algorithms in the literature. METHODS: Demand for medical expertise far outstrips supply, with tens of millions in the US alone with deficient access to specialty care. Rather than potentially months long delays to initiate diagnostic workup and medical treatment with a specialist, referring primary care supported by an automated recommender algorithm could anticipate and directly initiate patient evaluation that would otherwise be needed at subsequent a specialist appointment. We propose a novel graph representation learning approach with a heterogeneous graph neural network to model structured electronic health records and formulate recommendation/prediction of subsequent specialist orders as a link prediction problem. RESULTS: Models are trained and assessed in two specialty care sites: endocrinology and hematology. Our experimental results show that our model achieves an 8% improvement in ROC-AUC for endocrinology (ROC-AUC = 0.88) and 5% improvement for hematology (ROC-AUC = 0.84) personalized procedure recommendations over prior medical recommender systems. These recommender algorithm approaches provide medical procedure recommendations for endocrinology referrals more effectively than manual clinical checklists (recommender: precision = 0.60, recall = 0.27, F1-score = 0.37) vs. (checklist: precision = 0.16, recall = 0.28, F1-score = 0.20), and similarly for hematology referrals (recommender: precision = 0.44, recall = 0.38, F1-score = 0.41) vs. (checklist: precision = 0.27, recall = 0.71, F1-score = 0.39). CONCLUSION: Embedding graph neural network models into clinical care can improve digital specialty consultation systems and expand the access to medical experience of prior similar cases.

Large-scale physical activity data reveal worldwide activity inequality.

To be able to curb the global pandemic of physical inactivity and the associated 5.3 million deaths per year, we need to understand the basic principles that govern physical activity. However, there is a lack of large-scale measurements of physical activity patterns across free-living populations worldwide. Here we leverage the wide usage of smartphones with built-in accelerometry to measure physical activity at the global scale. We study a dataset consisting of 68 million days of physical activity for 717,527 people, giving us a window into activity in 111 countries across the globe. We find inequality in how activity is distributed within countries and that this inequality is a better predictor of obesity prevalence in the population than average activity volume. Reduced activity in females contributes to a large portion of the observed activity inequality. Aspects of the built environment, such as the walkability of a city, are associated with a smaller gender gap in activity and lower activity inequality. In more walkable cities, activity is greater throughout the day and throughout the week, across age, gender, and body mass index (BMI) groups, with the greatest increases in activity found for females. Our findings have implications for global public health policy and urban planning and highlight the role of activity inequality and the built environment in improving physical activity and health.

Evolution of resilience in protein interactomes across the tree of life.

Phenotype robustness to environmental fluctuations is a common biological phenomenon. Although most phenotypes involve multiple proteins that interact with each other, the basic principles of how such interactome networks respond to environmental unpredictability and change during evolution are largely unknown. Here we study interactomes of 1,840 species across the tree of life involving a total of 8,762,166 protein-protein interactions. Our study focuses on the resilience of interactomes to network failures and finds that interactomes become more resilient during evolution, meaning that interactomes become more robust to network failures over time. In bacteria, we find that a more resilient interactome is in turn associated with the greater ability of the organism to survive in a more complex, variable, and competitive environment. We find that at the protein family level proteins exhibit a coordinated rewiring of interactions over time and that a resilient interactome arises through gradual change of the network topology. Our findings have implications for understanding molecular network structure in the context of both evolution and environment.

Modeling polypharmacy side effects with graph convolutional networks.

Motivation: The use of drug combinations, termed polypharmacy, is common to treat patients with complex diseases or co-existing conditions. However, a major consequence of polypharmacy is a much higher risk of adverse side effects for the patient. Polypharmacy side effects emerge because of drug-drug interactions, in which activity of one drug may change, favorably or unfavorably, if taken with another drug. The knowledge of drug interactions is often limited because these complex relationships are rare, and are usually not observed in relatively small clinical testing. Discovering polypharmacy side effects thus remains an important challenge with significant implications for patient mortality and morbidity. Results: Here, we present Decagon, an approach for modeling polypharmacy side effects. The approach constructs a multimodal graph of protein-protein interactions, drug-protein target interactions and the polypharmacy side effects, which are represented as drug-drug interactions, where each side effect is an edge of a different type. Decagon is developed specifically to handle such multimodal graphs with a large number of edge types. Our approach develops a new graph convolutional neural network for multirelational link prediction in multimodal networks. Unlike approaches limited to predicting simple drug-drug interaction values, Decagon can predict the exact side effect, if any, through which a given drug combination manifests clinically. Decagon accurately predicts polypharmacy side effects, outperforming baselines by up to 69%. We find that it automatically learns representations of side effects indicative of co-occurrence of polypharmacy in patients. Furthermore, Decagon models particularly well polypharmacy side effects that have a strong molecular basis, while on predominantly non-molecular side effects, it achieves good performance because of effective sharing of model parameters across edge types. Decagon opens up opportunities to use large pharmacogenomic and patient population data to flag and prioritize polypharmacy side effects for follow-up analysis via formal pharmacological studies. Availability and implementation: Source code and preprocessed datasets are at: http://snap.stanford.edu/decagon.

Machine Learning for Integrating Data in Biology and Medicine: Principles, Practice, and Opportunities.

New technologies have enabled the investigation of biology and human health at an unprecedented scale and in multiple dimensions. These dimensions include myriad properties describing genome, epigenome, transcriptome, microbiome, phenotype, and lifestyle. No single data type, however, can capture the complexity of all the factors relevant to understanding a phenomenon such as a disease. Integrative methods that combine data from multiple technologies have thus emerged as critical statistical and computational approaches. The key challenge in developing such approaches is the identification of effective models to provide a comprehensive and relevant systems view. An ideal method can answer a biological or medical question, identifying important features and predicting outcomes, by harnessing heterogeneous data across several dimensions of biological variation. In this Review, we describe the principles of data integration and discuss current methods and available implementations. We provide examples of successful data integration in biology and medicine. Finally, we discuss current challenges in biomedical integrative methods and our perspective on the future development of the field.

Predicting multicellular function through multi-layer tissue networks.

MOTIVATION: Understanding functions of proteins in specific human tissues is essential for insights into disease diagnostics and therapeutics, yet prediction of tissue-specific cellular function remains a critical challenge for biomedicine. RESULTS: Here, we present OhmNet , a hierarchy-aware unsupervised node feature learning approach for multi-layer networks. We build a multi-layer network, where each layer represents molecular interactions in a different human tissue. OhmNet then automatically learns a mapping of proteins, represented as nodes, to a neural embedding-based low-dimensional space of features. OhmNet encourages sharing of similar features among proteins with similar network neighborhoods and among proteins activated in similar tissues. The algorithm generalizes prior work, which generally ignores relationships between tissues, by modeling tissue organization with a rich multiscale tissue hierarchy. We use OhmNet to study multicellular function in a multi-layer protein interaction network of 107 human tissues. In 48 tissues with known tissue-specific cellular functions, OhmNet provides more accurate predictions of cellular function than alternative approaches, and also generates more accurate hypotheses about tissue-specific protein actions. We show that taking into account the tissue hierarchy leads to improved predictive power. Remarkably, we also demonstrate that it is possible to leverage the tissue hierarchy in order to effectively transfer cellular functions to a functionally uncharacterized tissue. Overall, OhmNet moves from flat networks to multiscale models able to predict a range of phenotypes spanning cellular subsystems. AVAILABILITY AND IMPLEMENTATION: Source code and datasets are available at http://snap.stanford.edu/ohmnet . CONTACT: jure@cs.stanford.edu.

Analyzing Information Seeking and Drug-Safety Alert Response by Health Care Professionals as New Methods for Surveillance.

BACKGROUND: Patterns in general consumer online search logs have been used to monitor health conditions and to predict health-related activities, but the multiple contexts within which consumers perform online searches make significant associations difficult to interpret. Physician information-seeking behavior has typically been analyzed through survey-based approaches and literature reviews. Activity logs from health care professionals using online medical information resources are thus a valuable yet relatively untapped resource for large-scale medical surveillance. OBJECTIVE: To analyze health care professionals' information-seeking behavior and assess the feasibility of measuring drug-safety alert response from the usage logs of an online medical information resource. METHODS: Using two years (2011-2012) of usage logs from UpToDate, we measured the volume of searches related to medical conditions with significant burden in the United States, as well as the seasonal distribution of those searches. We quantified the relationship between searches and resulting page views. Using a large collection of online mainstream media articles and Web log posts we also characterized the uptake of a Food and Drug Administration (FDA) alert via changes in UpToDate search activity compared with general online media activity related to the subject of the alert. RESULTS: Diseases and symptoms dominate UpToDate searches. Some searches result in page views of only short duration, while others consistently result in longer-than-average page views. The response to an FDA alert for Celexa, characterized by a change in UpToDate search activity, differed considerably from general online media activity. Changes in search activity appeared later and persisted longer in UpToDate logs. The volume of searches and page view durations related to Celexa before the alert also differed from those after the alert. CONCLUSIONS: Understanding the information-seeking behavior associated with online evidence sources can offer insight into the information needs of health professionals and enable large-scale medical surveillance. Our Web log mining approach has the potential to monitor responses to FDA alerts at a national level. Our findings can also inform the design and content of evidence-based medical information resources such as UpToDate.

Predicting drug outcome of population via clinical knowledge graph.

Optimal treatments depend on numerous factors such as drug chemical properties, disease biology, and patient characteristics to which the treatment is applied. To realize the promise of AI in healthcare, there is a need for designing systems that can capture patient heterogeneity and relevant biomedical knowledge. Here we present PlaNet, a geometric deep learning framework that reasons over population variability, disease biology, and drug chemistry by representing knowledge in the form of a massive clinical knowledge graph that can be enhanced by language models. Our framework is applicable to any sub-population, any drug as well drug combinations, any disease, and to a wide range of pharmacological tasks. We apply the PlaNet framework to reason about outcomes of clinical trials: PlaNet predicts drug efficacy and adverse events, even for experimental drugs and their combinations that have never been seen by the model. Furthermore, PlaNet can estimate the effect of changing population on the trial outcome with direct implications on patient stratification in clinical trials. PlaNet takes fundamental steps towards AI-guided clinical trials design, offering valuable guidance for realizing the vision of precision medicine using AI.

Predicting transcriptional outcomes of novel multigene perturbations with GEARS.

Understanding cellular responses to genetic perturbation is central to numerous biomedical applications, from identifying genetic interactions involved in cancer to developing methods for regenerative medicine. However, the combinatorial explosion in the number of possible multigene perturbations severely limits experimental interrogation. Here, we present graph-enhanced gene activation and repression simulator (GEARS), a method that integrates deep learning with a knowledge graph of gene-gene relationships to predict transcriptional responses to both single and multigene perturbations using single-cell RNA-sequencing data from perturbational screens. GEARS is able to predict outcomes of perturbing combinations consisting of genes that were never experimentally perturbed. GEARS exhibited 40% higher precision than existing approaches in predicting four distinct genetic interaction subtypes in a combinatorial perturbation screen and identified the strongest interactions twice as well as prior approaches. Overall, GEARS can predict phenotypically distinct effects of multigene perturbations and thus guide the design of perturbational experiments.