Periodic Reporting for period 2 - REALMENT (Using real-world big data from eHealth, biobanks and national registries, integrated with clinical trial data to improve outcome of severe mental disorders)
Reporting period: 2022-12-01 to 2024-05-31
Mental disorders are amongst the largest disease groups and represent a major public health concern, posing a large financial burden on the European health care system. The total cost of mental disorders was recently estimated at more than 4% of GDP – or over €600 billion – in Europe. Overall, individuals with severe mental disorders have not benefited from the general advances in health care in the last decades, largely due to a lack of progress in pharmacological treatment and disease management. Moreover, the presence of multimorbidities, including both comorbid mental and somatic diseases, adds to the suffering and large reduction in quality of life (QoL) experienced by this patient group.
The main aim of REALMENT is to unleash the potential of clinical trial data in combination with large psychopharmacological Real-World Data (electronic Health Records, health registries, genotyped biobanks) applying novel Artificial Intelligence and Machine Learning technology to enable a personalized approach (‘precision psychiatry’) for clinical application in a Clinical Management Platform (4MENT).
We will exploit our unique access to large-scale available Real World Data samples derived from cohorts, medical records and the unique Nordic biobanks and registries, where lifespan diagnostic and prescription information is available from each individual. We will apply cutting-edge “big data” analytical approaches to Real World Data from Nordic, Baltic, British, Dutch and Italian cohorts to: i) identify and validate genetic markers of treatment outcome and multimorbidities, ii) develop tools suitable for prediction and stratification of treatment and iii) identify sub-populations that would benefit from preventive strategies.
The main aim of REALMENT is to unleash the potential of clinical trial data in combination with large psychopharmacological Real-World Data (electronic Health Records, health registries, genotyped biobanks) applying novel Artificial Intelligence and Machine Learning technology to enable a personalized approach (‘precision psychiatry’) for clinical application in a Clinical Management Platform (4MENT).
We will exploit our unique access to large-scale available Real World Data samples derived from cohorts, medical records and the unique Nordic biobanks and registries, where lifespan diagnostic and prescription information is available from each individual. We will apply cutting-edge “big data” analytical approaches to Real World Data from Nordic, Baltic, British, Dutch and Italian cohorts to: i) identify and validate genetic markers of treatment outcome and multimorbidities, ii) develop tools suitable for prediction and stratification of treatment and iii) identify sub-populations that would benefit from preventive strategies.
During the 2nd period, from 18 to 36 months, the REALMENT project has moved forward largely according to plan.
Real World Data: We continued to curate and harmonize multimorbidity and pharmacological data across cohorts for specific projects, see below.
Infrastructure and methods: We expanded our infrastructure for algorithm development based on partners secure data systems, and started adding Danish, Estonian and Swedish partners to the Dutch and Norwegian infrastructures. We coordinate software container solutions for integrated distributed analyses of Nordic data overall. The UK and Estonian partners have initiated initiatives on electronic data mining approaches.
Main results and achievements:
We published an overview article on how Real-World Data Can Facilitate the Development of Precision Medicine Treatment in Psychiatry.
Algorithm development: We have further tested machine learning (ML) models, such as regression and random forest algorithms, alongside deep learning models, specifically artificial feed forward neural networks, on both GWAS-informed and raw genotype patient-level data. We continue to work on the Finemap-MiXeR, which determines the causal single nucleotide polymorphisms (SNPs) associated with a trait at a given locus.
This work will form the basis for better models to predict treatment trajectories.
Gene discoveries: We expanded our previous GWAS on treatment response and meta-analysed with a Finnish sample. The results identified a single genome-wide significant locus tagging the GATA4 gene, which is described as a transcriptional regulator of genes involved in antipsychotic pharmacokinetics and thus a potentially actionable target. Work is still on-going to identify genetic variants associated with side-effects from SSRIs, involving several partners.
Genetics of trajectories: We initiated the work on the identification of genetic factors associated with treatment trajectories and are developing a framework to systematically evaluate the familial and genetic components influencing pharmacological trajectories. We saw an effect of age on onset on the probability of transitioning to clozapine, with an increased chance for early onset patients. Also, preliminary results have shown a male specific effect of family history of schizophrenia on the eventual prescription of clozapine. We are extending these analyses to other outcomes and medications while developing a framework to systematically evaluate the familial and genetic components influencing pharmacological trajectories.
Treatment response:
Prediction: We have worked on new approaches to increase the predictive power of personalized polygenic risk scores. We found that polygenic risk score for schizophrenia was significantly positively associated with prescribed (standardized) antipsychotic dose and antipsychotic polypharmacy with direction effects the same all cross-border five cohorts tested. Further, we have expanded the work to included antidepressants, and have now a paper in review describing these findings.
4MENT management platform: We further developed the 4MENT clinical management platform by improving the design based on stakeholder input and incorporate algorithms for multimodal prediction of treatment-resistance schizophrenia, including polygenic risk scores and clinical variables as input. We also initiated the ethical, data protection and legal framework for the platform design;
Cooordination, dissemination and exploitation: The 2nd stakeholder forum was held at the 3rd annual meeting of the project. Finally, the website has updated with lay summaries and the project is continuously being presented at several conferences and events including the world congress of psychiatric genetics, the European College Neuropsychopharmacology (ECNP) and Psychiatric Genetics Consortium (PGC) lab meeting. Flyers on the project have been distributed at several places.
Some deliverables have been postponed due to our initial issues hiring personnel. However, we have now the personnel available, and are catching up with the activities.
Real World Data: We continued to curate and harmonize multimorbidity and pharmacological data across cohorts for specific projects, see below.
Infrastructure and methods: We expanded our infrastructure for algorithm development based on partners secure data systems, and started adding Danish, Estonian and Swedish partners to the Dutch and Norwegian infrastructures. We coordinate software container solutions for integrated distributed analyses of Nordic data overall. The UK and Estonian partners have initiated initiatives on electronic data mining approaches.
Main results and achievements:
We published an overview article on how Real-World Data Can Facilitate the Development of Precision Medicine Treatment in Psychiatry.
Algorithm development: We have further tested machine learning (ML) models, such as regression and random forest algorithms, alongside deep learning models, specifically artificial feed forward neural networks, on both GWAS-informed and raw genotype patient-level data. We continue to work on the Finemap-MiXeR, which determines the causal single nucleotide polymorphisms (SNPs) associated with a trait at a given locus.
This work will form the basis for better models to predict treatment trajectories.
Gene discoveries: We expanded our previous GWAS on treatment response and meta-analysed with a Finnish sample. The results identified a single genome-wide significant locus tagging the GATA4 gene, which is described as a transcriptional regulator of genes involved in antipsychotic pharmacokinetics and thus a potentially actionable target. Work is still on-going to identify genetic variants associated with side-effects from SSRIs, involving several partners.
Genetics of trajectories: We initiated the work on the identification of genetic factors associated with treatment trajectories and are developing a framework to systematically evaluate the familial and genetic components influencing pharmacological trajectories. We saw an effect of age on onset on the probability of transitioning to clozapine, with an increased chance for early onset patients. Also, preliminary results have shown a male specific effect of family history of schizophrenia on the eventual prescription of clozapine. We are extending these analyses to other outcomes and medications while developing a framework to systematically evaluate the familial and genetic components influencing pharmacological trajectories.
Treatment response:
Prediction: We have worked on new approaches to increase the predictive power of personalized polygenic risk scores. We found that polygenic risk score for schizophrenia was significantly positively associated with prescribed (standardized) antipsychotic dose and antipsychotic polypharmacy with direction effects the same all cross-border five cohorts tested. Further, we have expanded the work to included antidepressants, and have now a paper in review describing these findings.
4MENT management platform: We further developed the 4MENT clinical management platform by improving the design based on stakeholder input and incorporate algorithms for multimodal prediction of treatment-resistance schizophrenia, including polygenic risk scores and clinical variables as input. We also initiated the ethical, data protection and legal framework for the platform design;
Cooordination, dissemination and exploitation: The 2nd stakeholder forum was held at the 3rd annual meeting of the project. Finally, the website has updated with lay summaries and the project is continuously being presented at several conferences and events including the world congress of psychiatric genetics, the European College Neuropsychopharmacology (ECNP) and Psychiatric Genetics Consortium (PGC) lab meeting. Flyers on the project have been distributed at several places.
Some deliverables have been postponed due to our initial issues hiring personnel. However, we have now the personnel available, and are catching up with the activities.
We expect to make large progress on identifying predictors of treatment response of major psychopharmacological agents applying real-world data from the large biobanks and health registry data. This will enable us to perform a series of GWAS of phenotypes (treatment outcome), not yet possible with other types of data. As an example, we have performed the first GWAS of treatment-resistant depression.
A long-range sequencing project of CYP450 genes is in planning taking advantage of several partners’ samples. We expect to find variants involved in drug metabolism in patients, useful for prediction of drug response and side effects.
REALMENT will continue to identify genetic variation associated with treatment outcome and multimorbidities building on development of suitable analytical tools, including prediction and stratification algorithms. We will validate our discoveries in independent samples, both Real-World Data and clinical trial data, and develop a management platform for improved outcome and quality of life for individuals with mental disorders.
A long-range sequencing project of CYP450 genes is in planning taking advantage of several partners’ samples. We expect to find variants involved in drug metabolism in patients, useful for prediction of drug response and side effects.
REALMENT will continue to identify genetic variation associated with treatment outcome and multimorbidities building on development of suitable analytical tools, including prediction and stratification algorithms. We will validate our discoveries in independent samples, both Real-World Data and clinical trial data, and develop a management platform for improved outcome and quality of life for individuals with mental disorders.