AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles (e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities, and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.
Based on the profile of genetic alterations occurring in tumor samples from selected diffuse large B-cell lymphoma (DLBCL) patients, 2 recent whole-exome sequencing studies proposed partially overlapping classification systems. Using clustering techniques applied to targeted sequencing data derived from a large unselected population-based patient cohort with full clinical follow-up (n = 928), we investigated whether molecular subtypes can be robustly identified using methods potentially applicable in routine clinical practice. DNA extracted from DLBCL tumors diagnosed in patients residing in a catchment population of ∼4 million (14 centers) were sequenced with a targeted 293-gene hematological-malignancy panel. Bernoulli mixture-model clustering was applied and the resulting subtypes analyzed in relation to their clinical characteristics and outcomes. Five molecular subtypes were resolved, termed MYD88, BCL2, SOCS1/SGK1, TET2/SGK1, and NOTCH2, along with an unclassified group. The subtypes characterized by genetic alterations of BCL2, NOTCH2, and MYD88 recapitulated recent studies showing good, intermediate, and poor prognosis, respectively. The SOCS1/SGK1 subtype showed biological overlap with primary mediastinal B-cell lymphoma and conferred excellent prognosis. Although not identified as a distinct cluster, NOTCH1 mutation was associated with poor prognosis. The impact of TP53 mutation varied with genomic subtypes, conferring no effect in the NOTCH2 subtype and poor prognosis in the MYD88 subtype. Our findings confirm the existence of molecular subtypes of DLBCL, providing evidence that genomic tests have prognostic significance in non-selected DLBCL patients. The identification of both good and poor risk subtypes in patients treated with R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) clearly show the clinical value of the approach, confirming the need for a consensus classification.
Precision health relies on the ability to assess disease risk at an individual level, detect early preclinical conditions and initiate preventive strategies. Recent technological advances in omics and wearable monitoring enable deep molecular and physiological profiling and may provide important tools for precision health. We explored the ability of deep longitudinal profiling to make health-related discoveries, identify clinically relevant molecular pathways, and impact behavior in a prospective longitudinal cohort ( n = 109) enriched for risk of type 2 diabetes mellitus (DM). The cohort underwent integrative Personalized Omics Profiling (iPOP) from samples collected quarterly for up to 8 years (median 2.8 years) using clinical measures and emerging technologies including genome, immunome, transcriptome, proteome, metabolome, microbiome, and wearable monitoring. We discovered over 67 clinically actionable health discoveries and identified multiple molecular pathways associated with metabolic, cardiovascular and oncologic pathophysiology. We developed prediction models for insulin resistance using omics measurements illustrating their potential to replace burdensome tests. Finally, study participation lead the majority of participants to implement diet and exercise changes. Altogether, we conclude that deep longitudinal profiling can lead to actionable health discoveries and provide relevant information for precision health.
Age-related macular degeneration (AMD) is thought to be the result of a lifetime of oxidative insult that results in photoreceptor death within the macula. Increased risk of AMD may result from low levels of lutein and zeaxanthin (macular pigment) in the diet, serum or retina, and excessive exposure to blue light. Through its light-screening capacity and antioxidant activity, macular pigment may reduce photooxidation in the central retina. Lutein supplements, at 30 mg/d, were shown previously to increase serum lutein and macular pigment density in two subjects. In this study, we compared the effects of a range of lutein doses (2.4- 30 mg/d), as well as a high zeaxanthin dose (30 mg/d), on the serum and macular pigment in a series of experiments. Serum carotenoids were quantified by HPLC. Macular pigment densities were determined psychophysically. Serum lutein concentrations in each subject reached a plateau that was correlated with the dose (r = 0.82, P < 0.001). Plateau concentrations ranged from 2.8 x 10(-7) to 2.7 x 10(-6) mol/L. Zeaxanthin was less well absorbed than an equal lutein dose, resulting in plateaus of approximately 5 x 10(-7) mol/L. The rate of increase in macular pigment optical density was correlated with the plateau concentration of carotenoids in the serum (r = 0.58, P < 0.001), but not with the presupplementation optical density (r = 0.13, P = 0.21). The mean rate of increase was (3.42 +/- 0.80) x 10(5) mAU/d per unit concentration (mol/L) of carotenoids in the serum. It remains to be demonstrated whether lutein or zeaxanthin dietary supplements reduce the incidence of AMD.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.