诺和诺德基金会

Investment in artificial intelligence (AI) has grown rapidly, yet its returns to scientific research remain poorly understood. We study how AI reshapes the production of science using a comprehensive dataset of research proposals submitted to a large international funding agency, including both funded and unfunded projects. Combining keyword extraction with large language model classification, we identify the presence, type, and functional role of AI within each proposal and link these measures to detailed budget allocations, team structure, and subsequent publication outcomes. We find that, in the short run, AI adoption is associated with modest improvements in scientific outcomes concentrated in the upper tail. Instead, its primary effects arise in the organization of research: AI-enabled projects reallocate resources toward human capital, involve larger teams, and undertake a broader set of tasks. These patterns are consistent with a reorganization of the scientific production process rather than immediate efficiency gains, in line with theories of general-purpose technologies. Task-level analyses further show that activities expanded in AI-enabled projects, particularly ideation and experimentation, are increasingly compatible with large language model capabilities, suggesting potential for future productivity gains as these technologies mature.

We consider the problem of indirect comparison, where a treatment arm of interest is absent by design in the target randomized control trial (RCT) but available in a source RCT. The identifiability of the target population average treatment effect often relies on conditional transportability assumptions. However, it is a common concern whether all relevant effect modifiers are measured and controlled for. We highlight a new proximal identification result in the presence of shifted, unobserved effect modifiers based on proxies: an adjustment proxy in both RCTs and an additional reweighting proxy in the source RCT. We propose an estimator which is doubly-robust against misspecifications of the so-called bridge functions and asymptotically normal under mild consistency of the nuisance models. An alternative estimator is presented to accommodate missing outcomes in the source RCT, which we then apply to conduct a proximal indirect comparison analysis using two weight management trials.

Smoking and obesity are associated with risk of hidradenitis suppurativa, and both are considered important environmental risk factors. However, a causal relationship remains unproven. To primarily investigate the relationship between body mass index (BMI, calculated as weight in kilograms divided by height in meters squared) and smoking and HS, and secondarily to investigate potential relationships between 3 inflammatory diseases (psoriasis, inflammatory bowel disease [IBD], and systemic sclerosis [SSc]) and HS. A mendelian randomization (MR) study conducted in 2024 on 5 exposure phenotypes (BMI, smoking, psoriasis, IBD, and SSc) on the outcome of phenotype HS was conducted. The MR analyses used large genetic White European cohorts from genome-wide association studies (GWAS) of each of the 6 phenotypes. Initial analyses were conducted May, 2024, and were updated in May, 2025. The 5 exposure phenotypes using predetermined genome-wide significant single-nucleotide variants as proxies for each particular exposure. The GWAS on HS included 4814 case patients and more than 1.2 million controls from Denmark, Iceland, Finland, the UK, and the US. The BMI GWAS involved 700 000 individuals from the UK Biobank and GIANT consortium. Smoking data were obtained from 1.23 million participants in an international consortium. The psoriasis GWAS analyzed 39 498 case patients and 286 769 controls from White European populations and a DNA genetic testing company. The IBD GWAS meta-analysis included 38 155 case patients and 48 485 controls from the International Inflammatory Bowel Disease (IBD) Genetics Consortium. The SSc GWAS included 9095 case patients and 17 584 controls from White European populations. Genetic correlations ( r g ) were found between HS and all exposure phenotypes except SSc (BMI: r g = 0.36, P < .001; smoking: r g = 0.33, P < .001; IBD: r g = 0.25, P < .001; psoriasis: r g = 0.34, P < .001; SSc: r g = 0.33, P = .22). MR analyses supported an effect of BMI on HS (β = 0.87; odds ratio [OR] per BMI unit, 1.20; 95% CI, 1.17-1.23; P < .001) without signs of pleiotropy (slope: β = 0.91, P < .001, P for intercept = .76). Smoking showed a significant causal estimate (β = 0.59, P < .001), but results became inconclusive in subsequent sensitivity analyses. Among IBD, psoriasis, and SSc, results supported a causal effect of IBD on HS (β = 0.18, OR = 1.20; 95% CI, 1.15-1.24; P < .001), without signs of pleiotropy. These findings indicate causal effects of IBD and increased BMI on the risk of HS. This information may help physicians inform patients about disease risk contributed by modifiable lifestyle behaviors, which can be beneficial for planning lifestyle interventions.

The aptamer-based SomaScan assay measures thousands of proteins. SomaLogic provides a multi-step pre-processing procedure to reduce the technical variability of this data. This paper will evaluate how each step of this procedure affects analysis results. We performed a comparative assessment using data from two randomised clinical trials in weight management. We show that SomaLogic’s adaptive normalization by maximum likelihood (ANML) procedure introduces a bias to fold change estimates, with a median bias of + 3.7% and + 3.4% in the two trials. The bias was confirmed by a simulation study, where ANML introduced false positive findings. Additionally, their plate scaling procedure has no effect on data when the calibration step is included. However, SomaLogic's pipeline excluding ANML does reduce technical variability without a substantial impact on fold change estimates. We recommend that researchers considering the use of ANML in clinical trials should verify the absence of this bias.