Periodic Reporting for period 2 - AI-PREVENT (A nationwide artificial intelligence risk assessment for primary prevention of cardiometabolic diseases)
Periodo di rendicontazione: 2022-07-01 al 2023-12-31
Diabetes, stroke and coronary artery disease (cardiometabolic diseases) are the leading cause of death in Europe.
Given that effective pharmacological and lifestyle interventions are available, it is important to identify high risk individuals at an early stage.
Traditionally, this is done using clinical prediction models.
However, the established models have substantial limitations: they are often used by doctors only when an underlying disease is already suspected, they are not developed on updated nationally-representative data, and they require time- consuming clinical measurements. T
hus, a substantial part of the population is not provided with risk assessment.
I propose to revolutionize the existing approaches to primary prevention by providing risk assessment of cardiometabolic diseases before an individual even steps into the doctor’s office for a visit.
To this end my project has three main objectives:
1) Development of artificial intelligence (AI) approaches to model health trajectories based on nationwide registry data on medications, diagnoses, familial risk and socio-demographic information to obtain accurate risk estimates for cardiometabolic disease.
I will integrate high quality data from selected countries that have long traditions of registry data (Finland and Sweden, over 7.5 million individuals).
2) Identification of health trajectories that maximize the clinical utility of genetic scores by integrating genetic and registry-based data on > 1 million people to identify subgroups of individuals for whom genetic information might improve risk prediction.
3) Validation of AI and genetic-based risk assessment as first-stage screening via a clinical study in 2800 individuals.
My project leverages the latest developments in AI and high-quality data of unprecedented scale to deliver a paradigm shift with important public health consequences by potentially changing the way cardiometabolic disease risk is assessed.
Given that effective pharmacological and lifestyle interventions are available, it is important to identify high risk individuals at an early stage.
Traditionally, this is done using clinical prediction models.
However, the established models have substantial limitations: they are often used by doctors only when an underlying disease is already suspected, they are not developed on updated nationally-representative data, and they require time- consuming clinical measurements. T
hus, a substantial part of the population is not provided with risk assessment.
I propose to revolutionize the existing approaches to primary prevention by providing risk assessment of cardiometabolic diseases before an individual even steps into the doctor’s office for a visit.
To this end my project has three main objectives:
1) Development of artificial intelligence (AI) approaches to model health trajectories based on nationwide registry data on medications, diagnoses, familial risk and socio-demographic information to obtain accurate risk estimates for cardiometabolic disease.
I will integrate high quality data from selected countries that have long traditions of registry data (Finland and Sweden, over 7.5 million individuals).
2) Identification of health trajectories that maximize the clinical utility of genetic scores by integrating genetic and registry-based data on > 1 million people to identify subgroups of individuals for whom genetic information might improve risk prediction.
3) Validation of AI and genetic-based risk assessment as first-stage screening via a clinical study in 2800 individuals.
My project leverages the latest developments in AI and high-quality data of unprecedented scale to deliver a paradigm shift with important public health consequences by potentially changing the way cardiometabolic disease risk is assessed.
The most important achievement for this project has been the establishment of the FinRegistry data resource (https://www.finregistry.fi/(si apre in una nuova finestra)).
FinRegistry is a curated, nationwide, register-based data resource for developing statistical and machine learning models, performing high-throughput epidemiological analyses and deriving outcome-specific prediction models.
FinRegistry data are collected across 19 registries covering public health care visits, health conditions, medications, vaccinations, laboratory responses, demographics, familial relations and socioeconomic variables, with decades of follow-up for most registries.
FinRegistry includes everyone living in Finland on 1 January 2010, as well as their parents, spouses, children and siblings, comprising a sample size of approximately 7.2 million persons.
FinRegistry data are mapped to more than 3000 clinical endpoints defined by leveraging multiple registers and clinical expertise as part of the FinnGen project.
The Risteys web portal [https://risteys.finregistry.fi] enables exploration of clinical endpoint definitions, their links to international ontologies and the results of epidemiological analyses to gain insights into disease epidemiology in the Finnish population.
The establishment of this resource and the integration of genetic data provided by FInnGen (https://www.finngen.fi/en(si apre in una nuova finestra)) has allowed to perform several machine learning analyses of direct public health relevance.
First, we considered the effects of 2,890 health, socio-economic and demographic factors in the entire Finnish population aged 30–80 and genome-wide information from 273,765 individuals to identify predictors of COVID-19 vaccination uptake. The strongest predictors of vaccination status were labour income and medication purchase history. Mental health conditions and having unvaccinated first-degree relatives were associated with reduced vaccination. A prediction model combining all predictors achieved good discrimination (area under the receiver operating characteristic curve, 0.801; 95% confidence interval, 0.799–0.803). The 1% of individuals with the highest predicted risk of not vaccinating had an observed vaccination rate of 18.8%, compared with 90.3% in the study population. We identified eight genetic loci associated with vaccination uptake and derived a polygenic score, which was a weak predictor in an independent subset. Our results suggest that individuals at higher risk of suffering the worst consequences of COVID-19 are also less likely to vaccinate.
Second, we studied the impact of genetic variation on overall disease burden. We introduce an approach to estimate the effect of genetic risk factors on disability-adjusted life years (DALYs; ‘lost healthy life years’). We use genetic information from 735,748 individuals and consider 80 diseases. Rare variants had the highest effect on DALYs at the individual level. Among common variants, rs3798220 (LPA) had the strongest individual-level effect, with 1.18 DALYs from carrying 1 versus 0 copies. Being in the top 10% versus the bottom 90% of a polygenic score for multisite chronic pain had an effect of 3.63 DALYs. Some common variants had a population-level effect comparable to modifiable risk factors such as high sodium intake and low physical activity. Attributable DALYs vary between males and females for some genetic exposures. We found that genetic risk factors can explain a sizable number of healthy life years lost both at the individual and population level.
These are the two main scientific output directly related to this project. We are now started to address the third Aim of the grant, focusing on the establishment of a clinical trial names GENEROOS. In this study, we will leverage the unique opportunity provided by the Finnish biobank research to re-contact 1200 individuals that have extreme genetic predisposition for high/low BMI as measured by a p[olygenic score for BMI and evaluate how a randomized diet intervention effect varies between the two extreme groups.
FinRegistry is a curated, nationwide, register-based data resource for developing statistical and machine learning models, performing high-throughput epidemiological analyses and deriving outcome-specific prediction models.
FinRegistry data are collected across 19 registries covering public health care visits, health conditions, medications, vaccinations, laboratory responses, demographics, familial relations and socioeconomic variables, with decades of follow-up for most registries.
FinRegistry includes everyone living in Finland on 1 January 2010, as well as their parents, spouses, children and siblings, comprising a sample size of approximately 7.2 million persons.
FinRegistry data are mapped to more than 3000 clinical endpoints defined by leveraging multiple registers and clinical expertise as part of the FinnGen project.
The Risteys web portal [https://risteys.finregistry.fi] enables exploration of clinical endpoint definitions, their links to international ontologies and the results of epidemiological analyses to gain insights into disease epidemiology in the Finnish population.
The establishment of this resource and the integration of genetic data provided by FInnGen (https://www.finngen.fi/en(si apre in una nuova finestra)) has allowed to perform several machine learning analyses of direct public health relevance.
First, we considered the effects of 2,890 health, socio-economic and demographic factors in the entire Finnish population aged 30–80 and genome-wide information from 273,765 individuals to identify predictors of COVID-19 vaccination uptake. The strongest predictors of vaccination status were labour income and medication purchase history. Mental health conditions and having unvaccinated first-degree relatives were associated with reduced vaccination. A prediction model combining all predictors achieved good discrimination (area under the receiver operating characteristic curve, 0.801; 95% confidence interval, 0.799–0.803). The 1% of individuals with the highest predicted risk of not vaccinating had an observed vaccination rate of 18.8%, compared with 90.3% in the study population. We identified eight genetic loci associated with vaccination uptake and derived a polygenic score, which was a weak predictor in an independent subset. Our results suggest that individuals at higher risk of suffering the worst consequences of COVID-19 are also less likely to vaccinate.
Second, we studied the impact of genetic variation on overall disease burden. We introduce an approach to estimate the effect of genetic risk factors on disability-adjusted life years (DALYs; ‘lost healthy life years’). We use genetic information from 735,748 individuals and consider 80 diseases. Rare variants had the highest effect on DALYs at the individual level. Among common variants, rs3798220 (LPA) had the strongest individual-level effect, with 1.18 DALYs from carrying 1 versus 0 copies. Being in the top 10% versus the bottom 90% of a polygenic score for multisite chronic pain had an effect of 3.63 DALYs. Some common variants had a population-level effect comparable to modifiable risk factors such as high sodium intake and low physical activity. Attributable DALYs vary between males and females for some genetic exposures. We found that genetic risk factors can explain a sizable number of healthy life years lost both at the individual and population level.
These are the two main scientific output directly related to this project. We are now started to address the third Aim of the grant, focusing on the establishment of a clinical trial names GENEROOS. In this study, we will leverage the unique opportunity provided by the Finnish biobank research to re-contact 1200 individuals that have extreme genetic predisposition for high/low BMI as measured by a p[olygenic score for BMI and evaluate how a randomized diet intervention effect varies between the two extreme groups.
It worth noticing that during the COVID-19 pandemic, our group has focused on addressing the challenges related to this diseases. We have established the COVID-19 host genetic initiative, which has been considered a breakthrough by the research community and the media. This initiative brings together the human genetics community to generate, share, and analyze data to learn the genetic determinants of COVID-19 susceptibility, severity, and outcomes. Such discoveries could help to identify individuals at unusually high or low risk, generate hypotheses for drug repurposing, and contribute to global knowledge of the biology of SARS-CoV-2 infection and disease. The initiative had three main goals:
1.Provide an environment to foster the sharing of resources to facilitate COVID-19 host genetics research (e.g. protocols, questionnaires).
2.Organize analytical activities across studies to identify genetic determinants of COVID-19 susceptibility and severity.
3.Provide a platform to share the results from such activities, as well as the individual-level data where possible, to benefit the broader scientific community.
1.Provide an environment to foster the sharing of resources to facilitate COVID-19 host genetics research (e.g. protocols, questionnaires).
2.Organize analytical activities across studies to identify genetic determinants of COVID-19 susceptibility and severity.
3.Provide a platform to share the results from such activities, as well as the individual-level data where possible, to benefit the broader scientific community.