Complex effects of sequence variants on lipid levels and coronary artery disease.

Many sequence variants have additive effects on blood lipid levels and, through that, on the risk of coronary artery disease (CAD). We show that variants also have non-additive effects and interact to affect lipid levels as well as affecting variance and correlations. Variance and correlation effects are often signatures of epistasis or gene-environmental interactions. These complex effects can translate into CAD risk. For example, Trp154Ter in FUT2 protects against CAD among subjects with the A1 blood group, whereas it associates with greater risk of CAD in others. His48Arg in ADH1B interacts with alcohol consumption to affect lipid levels and CAD. The effect of variants in TM6SF2 on blood lipids is greatest among those who never eat oily fish but absent from those who often do. This work demonstrates that variants that affect variance of quantitative traits can allow for the discovery of epistasis and interactions of variants with the environment.

Early divergent strains of Yersinia pestis in Eurasia 5,000 years ago.

The bacteria Yersinia pestis is the etiological agent of plague and has caused human pandemics with millions of deaths in historic times. How and when it originated remains contentious. Here, we report the oldest direct evidence of Yersinia pestis identified by ancient DNA in human teeth from Asia and Europe dating from 2,800 to 5,000 years ago. By sequencing the genomes, we find that these ancient plague strains are basal to all known Yersinia pestis. We find the origins of the Yersinia pestis lineage to be at least two times older than previous estimates. We also identify a temporal sequence of genetic changes that lead to increased virulence and the emergence of the bubonic plague. Our results show that plague infection was endemic in the human populations of Eurasia at least 3,000 years before any historical recordings of pandemics.

Immune Profiling of Human Gut-Associated Lymphoid Tissue Identifies a Role for Isolated Lymphoid Follicles in Priming of Region-Specific Immunity.

The intestine contains some of the most diverse and complex immune compartments in the body. Here we describe a method for isolating human gut-associated lymphoid tissues (GALTs) that allows unprecedented profiling of the adaptive immune system in submucosal and mucosal isolated lymphoid follicles (SM-ILFs and M-ILFs, respectively) as well as in GALT-free intestinal lamina propria (LP). SM-ILF and M-ILF showed distinct patterns of distribution along the length of the intestine, were linked to the systemic circulation through MAdCAM-1+ high endothelial venules and efferent lymphatics, and had immune profiles consistent with immune-inductive sites. IgA sequencing analysis indicated that human ILFs are sites where intestinal adaptive immune responses are initiated in an anatomically restricted manner. Our findings position ILFs as key inductive hubs for regional immunity in the human intestine, and the methods presented will allow future assessment of these compartments in health and disease.

A nondegenerate code of deleterious variants in Mendelian loci contributes to complex disease risk.

Although countless highly penetrant variants have been associated with Mendelian disorders, the genetic etiologies underlying complex diseases remain largely unresolved. By mining the medical records of over 110 million patients, we examine the extent to which Mendelian variation contributes to complex disease risk. We detect thousands of associations between Mendelian and complex diseases, revealing a nondegenerate, phenotypic code that links each complex disorder to a unique collection of Mendelian loci. Using genome-wide association results, we demonstrate that common variants associated with complex diseases are enriched in the genes indicated by this "Mendelian code." Finally, we detect hundreds of comorbidity associations among Mendelian disorders, and we use probabilistic genetic modeling to demonstrate that Mendelian variants likely contribute nonadditively to the risk for a subset of complex diseases. Overall, this study illustrates a complementary approach for mapping complex disease loci and provides unique predictions concerning the etiologies of specific diseases.

The sequences of 150,119 genomes in the UK Biobank.

Detailed knowledge of how diversity in the sequence of the human genome affects phenotypic diversity depends on a comprehensive and reliable characterization of both sequences and phenotypic variation. Over the past decade, insights into this relationship have been obtained from whole-exome sequencing or whole-genome sequencing of large cohorts with rich phenotypic data1,2. Here we describe the analysis of whole-genome sequencing of 150,119 individuals from the UK Biobank3. This constitutes a set of high-quality variants, including 585,040,410 single-nucleotide polymorphisms, representing 7.0% of all possible human single-nucleotide polymorphisms, and 58,707,036 indels. This large set of variants allows us to characterize selection based on sequence variation within a population through a depletion rank score of windows along the genome. Depletion rank analysis shows that coding exons represent a small fraction of regions in the genome subject to strong sequence conservation. We define three cohorts within the UK Biobank: a large British Irish cohort, a smaller African cohort and a South Asian cohort. A haplotype reference panel is provided that allows reliable imputation of most variants carried by three or more sequenced individuals. We identified 895,055 structural variants and 2,536,688 microsatellites, groups of variants typically excluded from large-scale whole-genome sequencing studies. Using this formidable new resource, we provide several examples of trait associations for rare variants with large effects not found previously through studies based on whole-exome sequencing and/or imputation.

Lifestyle factors in the biomedical literature: An ontology and comprehensive resources for named entity recognition.

MOTIVATION: Despite lifestyle factors (LSFs) being increasingly acknowledged in shaping individual health trajectories, particularly in chronic diseases, they have still not been systematically described in the biomedical literature. This is in part because no named entity recognition (NER) system exists, which can comprehensively detect all types of LSFs in text. The task is challenging due to their inherent diversity, lack of a comprehensive LSF classification for dictionary-based NER, and lack of a corpus for deep learning-based NER. RESULTS: We present a novel Lifestyle Factor Ontology (LSFO), which we used to develop a dictionary-based system for recognition and normalization of LSFs. Additionally, we introduce a manually annotated corpus for LSFs (LSF200) suitable for training and evaluation of NER systems, and use it to train a transformer-based system. Evaluating the performance of both NER systems on the corpus revealed an F-score of 64% for the dictionary-based system and 76% for the transformer-based system. Large-scale application of these systems on PubMed abstracts and PMC Open Access articles identified over 300 million mentions of LSF in the biomedical literature. AVAILABILITY: LSFO, the annotated LSF200 corpus, and the detected LSFs in PubMed and PMC-OA articles using both NER systems, are available under open licenses via the following GitHub repository: Https://github.com/EsmaeilNourani/LSFO-expansion. This repository contains links to two associated GitHub repositories and a Zenodo project related to the study. LSFO is also available at BioPortal: Https://bioportal.bioontology.org/ontologies/LSFO. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Leveraging European infrastructures to access 1 million human genomes by 2022.

Human genomics is undergoing a step change from being a predominantly research-driven activity to one driven through health care as many countries in Europe now have nascent precision medicine programmes. To maximize the value of the genomic data generated, these data will need to be shared between institutions and across countries. In recognition of this challenge, 21 European countries recently signed a declaration to transnationally share data on at least 1 million human genomes by 2022. In this Roadmap, we identify the challenges of data sharing across borders and demonstrate that European research infrastructures are well-positioned to support the rapid implementation of widespread genomic data access.

Author Correction: Leveraging European infrastructures to access 1 million human genomes by 2022.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Deep integrative models for large-scale human genomics.

Polygenic risk scores (PRSs) are expected to play a critical role in precision medicine. Currently, PRS predictors are generally based on linear models using summary statistics, and more recently individual-level data. However, these predictors mainly capture additive relationships and are limited in data modalities they can use. We developed a deep learning framework (EIR) for PRS prediction which includes a model, genome-local-net (GLN), specifically designed for large-scale genomics data. The framework supports multi-task learning, automatic integration of other clinical and biochemical data, and model explainability. When applied to individual-level data from the UK Biobank, the GLN model demonstrated a competitive performance compared to established neural network architectures, particularly for certain traits, showcasing its potential in modeling complex genetic relationships. Furthermore, the GLN model outperformed linear PRS methods for Type 1 Diabetes, likely due to modeling non-additive genetic effects and epistasis. This was supported by our identification of widespread non-additive genetic effects and epistasis in the context of T1D. Finally, we constructed PRS models that integrated genotype, blood, urine, and anthropometric data and found that this improved performance for 93% of the 290 diseases and disorders considered. EIR is available at https://github.com/arnor-sigurdsson/EIR.

A partial loss-of-function variant in STAT6 protects against T2 asthma.

BACKGROUND: Signal Transducer and Activator of Transcription 6 (STAT6) is central to Type 2 (T2) inflammation and common non-coding variants at the STAT6 locus associate with various T2 inflammatory traits, including diseases, and its pathway is widely targeted in asthma treatment. OBJECTIVE: To test the association of a rare missense variant in STAT6, p.L406P, with T2 inflammatory traits, including the risk of asthma and allergic diseases, and to characterize its functional consequences in cell culture. METHODS: We tested association of p.L406P with plasma protein levels, white blood cell counts and the risk of asthma and allergic phenotypes. We tested significant associations in other cohorts using a burden test. The effects of p.L406P on STAT6 protein function were examined in cell lines and by comparing CD4+ T-cell responses from carriers and non-carriers of the variant. RESULTS: p.L406P associated with reduced plasma levels of STAT6 and IgE as well as with lower eosinophil and basophil counts in blood. It also protected against asthma, mostly driven by severe T2 high asthma. We showed that p.L406P led to lower IL-4-induced activation in luciferase reporter assays and lower levels of STAT6 in CD4+ T cells. We identified multiple genes with expression that was affected by the p.L406P genotype upon IL-4 treatment of CD4+ T cells; the effect was consistent with a weaker IL-4 response in carriers than non-carriers of p.L406P. CONCLUSIONS: We report a partial loss-of-function variant in STAT6, resulting in dampened IL-4 responses and protection from T2 high asthma, implicating STAT6 as an attractive therapeutic target.

Gut microbiome shifts in people with type 1 diabetes are associated with glycaemic control: an INNODIA study.

AIMS/HYPOTHESIS: The gut microbiome is implicated in the disease process leading to clinical type 1 diabetes, but less is known about potential changes in the gut microbiome after the diagnosis of type 1 diabetes and implications in glucose homeostasis. We aimed to analyse potential associations between the gut microbiome composition and clinical and laboratory data during a 2 year follow-up of people with newly diagnosed type 1 diabetes, recruited to the Innovative approaches to understanding and arresting type 1 diabetes (INNODIA) study. In addition, we analysed the microbiome composition in initially unaffected family members, who progressed to clinical type 1 diabetes during or after their follow-up for 4 years. METHODS: We characterised the gut microbiome composition of 98 individuals with newly diagnosed type 1 diabetes (ND cohort) and 194 autoantibody-positive unaffected family members (UFM cohort), representing a subgroup of the INNODIA Natural History Study, using metagenomic sequencing. Participants from the ND cohort attended study visits within 6 weeks from the diagnosis and 3, 6, 12 and 24 months later for stool sample collection and laboratory tests (HbA1c, C-peptide, diabetes-associated autoantibodies). Participants from the UFM cohort were assessed at baseline and 6, 12, 18, 24 and 36 months later. RESULTS: We observed a longitudinal increase in 21 bacterial species in the ND cohort but not in the UFM cohort. The relative abundance of Faecalibacterium prausnitzii was inversely associated with the HbA1c levels at diagnosis (p=0.0019). The rate of the subsequent disease progression in the ND cohort, as assessed by change in HbA1c, C-peptide levels and insulin dose, was associated with the abundance of several bacterial species. Individuals with rapid decrease in C-peptide levels in the ND cohort had the lowest gut microbiome diversity. Nineteen individuals who were diagnosed with type 1 diabetes in the UFM cohort had increased abundance of Sutterella sp. KLE1602 compared with the undiagnosed UFM individuals (p=1.2 × 10-4). CONCLUSIONS/INTERPRETATION: Our data revealed associations between the gut microbiome composition and the disease progression in individuals with recent-onset type 1 diabetes. Future mechanistic studies as well as animal studies and human trials are needed to further validate the significance and causality of these associations.

Determinants of plasma levels of proglucagon and the metabolic impact of glucagon receptor signalling: a UK Biobank study.

AIMS/HYPOTHESES: Glucagon and glucagon-like peptide-1 (GLP-1) are derived from the same precursor; proglucagon, and dual agonists of their receptors are currently being explored for the treatment of obesity and metabolic dysfunction-associated steatotic liver disease (MASLD). Elevated levels of endogenous glucagon (hyperglucagonaemia) have been linked with hyperglycaemia in individuals with type 2 diabetes but are also observed in individuals with obesity and MASLD. GLP-1 levels have been reported to be largely unaffected or even reduced in similar conditions. We investigated potential determinants of plasma proglucagon and associations of glucagon receptor signalling with metabolic diseases based on data from the UK Biobank. METHODS: We used exome sequencing data from the UK Biobank for ~410,000 white participants to identify glucagon receptor variants and grouped them based on their known or predicted signalling. Data on plasma levels of proglucagon estimated using Olink technology were available for a subset of the cohort (~40,000). We determined associations of glucagon receptor variants and proglucagon with BMI, type 2 diabetes and liver fat (quantified by liver MRI) and performed survival analyses to investigate if elevated proglucagon predicts type 2 diabetes development. RESULTS: Obesity, MASLD and type 2 diabetes were associated with elevated plasma levels of proglucagon independently of each other. Baseline proglucagon levels were associated with the risk of type 2 diabetes development over a 14 year follow-up period (HR 1.13; 95% CI 1.09, 1.17; n=1562; p=1.3×10-12). This association was of the same magnitude across strata of BMI. Carriers of glucagon receptor variants with reduced cAMP signalling had elevated levels of proglucagon (ß 0.847; 95% CI 0.04, 1.66; n=17; p=0.04), and carriers of variants with a predicted frameshift mutation had higher levels of liver fat compared with the wild-type reference group (ß 0.504; 95% CI 0.03, 0.98; n=11; p=0.04). CONCLUSIONS/INTERPRETATION: Our findings support the suggestion that glucagon receptor signalling is involved in MASLD, that plasma levels of proglucagon are linked to the risk of type 2 diabetes development, and that proglucagon levels are influenced by genetic variation in the glucagon receptor, obesity, type 2 diabetes and MASLD. Determining the molecular signalling pathways downstream of glucagon receptor activation may guide the development of biased GLP-1/glucagon co-agonist with improved metabolic benefits. DATA AVAILABILITY: All coding is available through https://github.com/nicwin98/UK-Biobank-GCG.

Longitudinal metabolite and protein trajectories prior to diabetes mellitus diagnosis in Danish blood donors: a nested case-control study.

AIMS/HYPOTHESIS: Metabolic risk factors and plasma biomarkers for diabetes have previously been shown to change prior to a clinical diabetes diagnosis. However, these markers only cover a small subset of molecular biomarkers linked to the disease. In this study, we aimed to profile a more comprehensive set of molecular biomarkers and explore their temporal association with incident diabetes. METHODS: We performed a targeted analysis of 54 proteins and 171 metabolites and lipoprotein particles measured in three sequential samples spanning up to 11 years of follow-up in 324 individuals with incident diabetes and 359 individuals without diabetes in the Danish Blood Donor Study (DBDS) matched for sex and birth year distribution. We used linear mixed-effects models to identify temporal changes before a diabetes diagnosis, either for any incident diabetes diagnosis or for type 1 and type 2 diabetes mellitus diagnoses specifically. We further performed linear and non-linear feature selection, adding 28 polygenic risk scores to the biomarker pool. We tested the time-to-event prediction gain of the biomarkers with the highest variable importance, compared with selected clinical covariates and plasma glucose. RESULTS: We identified two proteins and 16 metabolites and lipoprotein particles whose levels changed temporally before diabetes diagnosis and for which the estimated marginal means were significant after FDR adjustment. Sixteen of these have not previously been described. Additionally, 75 biomarkers were consistently higher or lower in the years before a diabetes diagnosis. We identified a single temporal biomarker for type 1 diabetes, IL-17A/F, a cytokine that is associated with multiple other autoimmune diseases. Inclusion of 12 biomarkers improved the 10-year prediction of a diabetes diagnosis (i.e. the area under the receiver operating curve increased from 0.79 to 0.84), compared with clinical information and plasma glucose alone. CONCLUSIONS/INTERPRETATION: Systemic molecular changes manifest in plasma several years before a diabetes diagnosis. A particular subset of biomarkers shows distinct, time-dependent patterns, offering potential as predictive markers for diabetes onset. Notably, these biomarkers show shared and distinct patterns between type 1 diabetes and type 2 diabetes. After independent replication, our findings may be used to develop new clinical prediction models.

Role of human plasma metabolites in prediabetes and type 2 diabetes from the IMI-DIRECT study.

AIMS/HYPOTHESIS: Type 2 diabetes is a chronic condition that is caused by hyperglycaemia. Our aim was to characterise the metabolomics to find their association with the glycaemic spectrum and find a causal relationship between metabolites and type 2 diabetes. METHODS: As part of the Innovative Medicines Initiative - Diabetes Research on Patient Stratification (IMI-DIRECT) consortium, 3000 plasma samples were measured with the Biocrates AbsoluteIDQ p150 Kit and Metabolon analytics. A total of 911 metabolites (132 targeted metabolomics, 779 untargeted metabolomics) passed the quality control. Multivariable linear and logistic regression analysis estimates were calculated from the concentration/peak areas of each metabolite as an explanatory variable and the glycaemic status as a dependent variable. This analysis was adjusted for age, sex, BMI, study centre in the basic model, and additionally for alcohol, smoking, BP, fasting HDL-cholesterol and fasting triacylglycerol in the full model. Statistical significance was Bonferroni corrected throughout. Beyond associations, we investigated the mediation effect and causal effects for which causal mediation test and two-sample Mendelian randomisation (2SMR) methods were used, respectively. RESULTS: In the targeted metabolomics, we observed four (15), 34 (99) and 50 (108) metabolites (number of metabolites observed in untargeted metabolomics appear in parentheses) that were significantly different when comparing normal glucose regulation vs impaired glucose regulation/prediabetes, normal glucose regulation vs type 2 diabetes, and impaired glucose regulation vs type 2 diabetes, respectively. Significant metabolites were mainly branched-chain amino acids (BCAAs), with some derivatised BCAAs, lipids, xenobiotics and a few unknowns. Metabolites such as lysophosphatidylcholine a C17:0, sum of hexoses, amino acids from BCAA metabolism (including leucine, isoleucine, valine, N-lactoylvaline, N-lactoylleucine and formiminoglutamate) and lactate, as well as an unknown metabolite (X-24295), were associated with HbA1c progression rate and were significant mediators of type 2 diabetes from baseline to 18 and 48 months of follow-up. 2SMR was used to estimate the causal effect of an exposure on an outcome using summary statistics from UK Biobank genome-wide association studies. We found that type 2 diabetes had a causal effect on the levels of three metabolites (hexose, glutamate and caproate [fatty acid (FA) 6:0]), whereas lipids such as specific phosphatidylcholines (PCs) (namely PC aa C36:2, PC aa C36:5, PC ae C36:3 and PC ae C34:3) as well as the two n-3 fatty acids stearidonate (18:4n3) and docosapentaenoate (22:5n3) potentially had a causal role in the development of type 2 diabetes. CONCLUSIONS/INTERPRETATION: Our findings identify known BCAAs and lipids, along with novel N-lactoyl-amino acid metabolites, significantly associated with prediabetes and diabetes, that mediate the effect of diabetes from baseline to follow-up (18 and 48 months). Causal inference using genetic variants shows the role of lipid metabolism and n-3 fatty acids as being causal for metabolite-to-type 2 diabetes whereas the sum of hexoses is causal for type 2 diabetes-to-metabolite. Identified metabolite markers are useful for stratifying individuals based on their risk progression and should enable targeted interventions.

Intervening on the storage time of RBC units and its effects on adverse recipient outcomes using real-world data.

Randomized controlled trials (RCTs) have found no evidence that the storage time of transfused red blood cell (RBC) units affects recipient survival. However, inherent difficulties in conducting RBC transfusion RCTs have prompted critique of their design, analyses, and interpretation. Here, we address these issues by emulating hypothetical randomized trials using large real-world data to further clarify the adverse effects of storage time. We estimated the comparative effect of transfusing exclusively older vs fresher RBC units on the primary outcome of death, and the secondary composite end point of thromboembolic events, or death, using inverse probability weighting. Thresholds were defined as 1, 2, 3, and 4 weeks of storage. A large Danish blood transfusion database from the period 2008 to 2018 comprising >900 000 transfusion events defined the observational data. A total of 89 799 patients receiving >340 000 RBC transfusions during 28 days of follow-up met the eligibility criteria. Treatment with RBC units exclusively fresher than 1, 2, 3, and 4 weeks of storage was found to decrease the 28-day recipient mortality with 2.44 percentage points (pp) (0.86 pp, 4.02 pp), 1.93 pp (0.85 pp, 3.02 pp), 1.06 pp (-0.20 pp, 2.33 pp), and -0.26 pp (-1.78 pp, 1.25 pp) compared with transfusing exclusively older RBC units, respectively. The 28-day risk differences for the composite end point were similar. This study suggests that transfusing exclusively older RBC units stored for >1 or 2 weeks increases the 28-day recipient mortality and risk of thromboembolism or death compared with transfusing fresher RBC units.

Multi-omics analysis reveals drivers of loss of ß-cell function after newly diagnosed autoimmune type 1 diabetes: An INNODIA multicenter study.

AIMS: Heterogeneity in the rate of ß-cell loss in newly diagnosed type 1 diabetes patients is poorly understood and creates a barrier to designing and interpreting disease-modifying clinical trials. Integrative analyses of baseline multi-omics data obtained after the diagnosis of type 1 diabetes may provide mechanistic insight into the diverse rates of disease progression after type 1 diabetes diagnosis. METHODS: We collected samples in a pan-European consortium that enabled the concerted analysis of five different omics modalities in data from 97 newly diagnosed patients. In this study, we used Multi-Omics Factor Analysis to identify molecular signatures correlating with post-diagnosis decline in ß-cell mass measured as fasting C-peptide. RESULTS: Two molecular signatures were significantly correlated with fasting C-peptide levels. One signature showed a correlation to neutrophil degranulation, cytokine signalling, lymphoid and non-lymphoid cell interactions and G-protein coupled receptor signalling events that were inversely associated with a rapid decline in ß-cell function. The second signature was related to translation and viral infection was inversely associated with change in ß-cell function. In addition, the immunomics data revealed a Natural Killer cell signature associated with rapid ß-cell decline. CONCLUSIONS: Features that differ between individuals with slow and rapid decline in ß-cell mass could be valuable in staging and prediction of the rate of disease progression and thus enable smarter (shorter and smaller) trial designs for disease modifying therapies as well as offering biomarkers of therapeutic effect.

BALDR: A Web-based platform for informed comparison and prioritization of biomarker candidates for type 2 diabetes mellitus.

Novel biomarkers are key to addressing the ongoing pandemic of type 2 diabetes mellitus. While new technologies have improved the potential of identifying such biomarkers, at the same time there is an increasing need for informed prioritization to ensure efficient downstream verification. We have built BALDR, an automated pipeline for biomarker comparison and prioritization in the context of diabetes. BALDR includes protein, gene, and disease data from major public repositories, text-mining data, and human and mouse experimental data from the IMI2 RHAPSODY consortium. These data are provided as easy-to-read figures and tables enabling direct comparison of up to 20 biomarker candidates for diabetes through the public website https://baldr.cpr.ku.dk.

No association between migraine and HLA alleles in a cohort of 13,210 individuals with migraine from the Danish Blood Donor Study.

OBJECTIVE: To determine the association between human leukocyte antigen (HLA) alleles and migraine, migraine subtypes, and sex-specific factors. BACKGROUND: It has long been hypothesized that inflammation contributes to migraine pathophysiology. This study examined the association between migraine and alleles in the HLA system, a key player in immune response and genetic diversity. METHODS: We performed a case-control study and included 13,210 individuals with migraine and 86,738 controls. All participants were part of the Danish Blood Donor Study Genomic Cohort. Participants were genotyped and 111 HLA alleles on 15 HLA genes were imputed. We examined the association between HLA alleles and migraine subtypes, considering sex-specific differences. RESULTS: We found no association between HLA alleles and migraine, neither overall, nor in the sex-specific analysis. In the migraine subtype analysis, three HLA alleles were associated with migraine without aura; however, these associations could not be replicated in an independent Icelandic cohort (2191 individuals with migraine without aura and 278,858 controls). Furthermore, we found no association between HLA alleles and migraine with aura or chronic migraine. CONCLUSION: We found no evidence of an association between the HLA system and migraine, suggesting that genetic factors related to the HLA system do not play a significant role in migraine susceptibility.

Adopting Mechanistic Molecular Biology Approaches in Exposome Research for Causal Understanding.

Through investigating the combined impact of the environmental exposures experienced by an individual throughout their lifetime, exposome research provides opportunities to understand and mitigate negative health outcomes. While current exposome research is driven by epidemiological studies that identify associations between exposures and effects, new frameworks integrating more substantial population-level metadata, including electronic health and administrative records, will shed further light on characterizing environmental exposure risks. Molecular biology offers methods and concepts to study the biological and health impacts of exposomes in experimental and computational systems. Of particular importance is the growing use of omics readouts in epidemiological and clinical studies. This paper calls for the adoption of mechanistic molecular biology approaches in exposome research as an essential step in understanding the genotype and exposure interactions underlying human phenotypes. A series of recommendations are presented to make the necessary and appropriate steps to move from exposure association to causation, with a huge potential to inform precision medicine and population health. This includes establishing hypothesis-driven laboratory testing within the exposome field, supported by appropriate methods to read across from model systems research to human.

Seasonally adjusted laboratory reference intervals to improve the performance of machine learning models for classification of cardiovascular diseases.

BACKGROUND: Variation in laboratory healthcare data due to seasonal changes is a widely accepted phenomenon. Seasonal variation is generally not systematically accounted for in healthcare settings. This study applies a newly developed adjustment method for seasonal variation to analyze the effect seasonality has on machine learning model classification of diagnoses. METHODS: Machine learning methods were trained and tested on ~ 22 million unique records from ~ 575,000 unique patients admitted to Danish hospitals. Four machine learning models (adaBoost, decision tree, neural net, and random forest) classifying 35 diseases of the circulatory system (ICD-10 diagnosis codes, chapter IX) were run before and after seasonal adjustment of 23 laboratory reference intervals (RIs). The effect of the adjustment was benchmarked via its contribution to machine learning models trained using hyperparameter optimization and assessed quantitatively using performance metrics (AUROC and AUPRC). RESULTS: Seasonally adjusted RIs significantly improved cardiovascular disease classification in 24 of the 35 tested cases when using neural net models. Features with the highest average feature importance (via SHAP explainability) across all disease models were sex, C- reactive protein, and estimated glomerular filtration. Classification of diseases of the vessels, such as thrombotic diseases and other atherosclerotic diseases consistently improved after seasonal adjustment. CONCLUSIONS: As data volumes increase and data-driven methods are becoming more advanced, it is essential to improve data quality at the pre-processing level. This study presents a method that makes it feasible to introduce seasonally adjusted RIs into the clinical research space in any disease domain. Seasonally adjusted RIs generally improve diagnoses classification and thus, ought to be considered and adjusted for in clinical decision support methods.