Periodic Reporting for period 1 - SUBTREAT (Subtype as a key to reduce heterogeneity of treatment effects in major depressive disorder)
Okres sprawozdawczy: 2022-10-01 do 2025-03-31
Major depressive disorder (MDD) is a leading contributor to disability and suicide. It is the most costly brain disorder in Europe. Although multiple treatments are of proven efficacy, individual responses to treatments vary considerably and MDD recurrence is common. There is considerable motivation to improve treatment regimen for individuals with MDD. However, it has been challenging because of the fundamental lack of understanding about the causes of variable treatment outcomes. MDD is widely accepted as a heterogeneous disorder; yet, most research strategies effectively consider MDD as a single disorder. Progress in understanding the variable treatment response will depend on “patient stratification”, i.e. identifying and accounting for patient heterogeneity when evaluating treatment efficacy.
SUBTREAT proposes a unique direction which considers subtype as the key to link aetiological and treatment effect heterogeneity. Our approach is to break down the heterogeneous treatment outcomes of MDD into more narrowly defined subtypes with divergent aetiologies. Specially, I propose three work packages: 1) dissect treatment heterogeneity across subtypes; a particularly innovative aspect of SUBTREAT is that we will use advanced data science approaches to identify novel subtypes which correlate with differential treatment outcomes; 2) determine divergent causes underlying MDD subtypes; we will comprehensively investigate causes at three levels including genetic and causal epidemiological risk factors, and brain cell types; 3) develop a novel prediction algorithm for treatment outcomes stratified by patient subgroups. SUBTREAT will illuminate the causes of MDD subtypes and the principal patterns of how subtypes contribute to differential long-term treatment outcomes. SUBTREAT findings will promote targeted drug development and treatment optimization for patient subgroups to achieve precision psychiatry.
SUBTREAT proposes a unique direction which considers subtype as the key to link aetiological and treatment effect heterogeneity. Our approach is to break down the heterogeneous treatment outcomes of MDD into more narrowly defined subtypes with divergent aetiologies. Specially, I propose three work packages: 1) dissect treatment heterogeneity across subtypes; a particularly innovative aspect of SUBTREAT is that we will use advanced data science approaches to identify novel subtypes which correlate with differential treatment outcomes; 2) determine divergent causes underlying MDD subtypes; we will comprehensively investigate causes at three levels including genetic and causal epidemiological risk factors, and brain cell types; 3) develop a novel prediction algorithm for treatment outcomes stratified by patient subgroups. SUBTREAT will illuminate the causes of MDD subtypes and the principal patterns of how subtypes contribute to differential long-term treatment outcomes. SUBTREAT findings will promote targeted drug development and treatment optimization for patient subgroups to achieve precision psychiatry.
The primary objective of SUBTREAT is to illuminate the diverse causes of major depressive disorder (MDD) subtypes and the principal patterns of how subtypes contribute to differential long-term treatment outcomes. The research activities outlined in the Description of the Action are well under way. We have already had several key publications during the first 24 months. We will describe the research and technological achievements within each work package:
Work Package 1: Dissecting MDD Subtypes in Relation to Long-Term Treatment Outcomes
This work package focuses on examining the differential treatment outcomes in clinically-informed MDD subtypes and identifying novel, data-driven subtypes using Nordic register/EMR data.
Objective 1.1. Assess and compare long-term treatment outcomes in clinically-informed subtypes
To successfully complete this objective, we have reviewed the literature and have had multiple rounds of discussions with clinical psychiatrists regarding important outcomes relevant to MDD. We have now completed analyses comparing these outcomes at the population level and within siblings. Observational studies are often subject to the confounding factors. By comparing the associations at these two levels help to draw conclusions that are closer to causal inference. The findings are clear that more severe subtypes (e.g. severe MDD) have much worse treatment outcomes than less severe ones (e.g. mild-moderate MDD). We are currently preparing a manuscript (Tong et al, in prep).
Objective 1.2. Identify data-driven subtypes that differ in treatment outcomes
We used comprehensive longitudinal data with over ten years of follow-up to assess treatment density, hospitalizations, recurrence, chronicity, diagnostic conversion, treatment resistance, disability, suicide, and premature mortality. A key achievement is the identification of new, data-driven MDD subtypes that offer more nuanced insights into treatment outcomes than the clinically informed subtypes. We further confirmed that these data-driven subtypes are replicable and genetically validated using family history measures. We are currently preparing a manuscript (Lu et al, in prep).
Work Package 2: Determining the Genetic and Epidemiological Causes Underlying MDD Subtypes
Objective 2.1. identify specific genetic risk factors for MDD subtypes
We are currently leading several large-scale genome-wide association studies (GWAS), identifying specific genetic loci associated with different MDD subtypes, including early-onset and late-onset MDD (Shorter et al, under review Nature Genetics), treatment-resistant MDD (Xiong et al, revision submitted to Molecular Psychiatry), and atypical MDD (Marshall et al, in prep).
Objective 2.2 test the causality of epidemiological risk factors in each subtype
The application of Mendelian randomization allowed us to identify causal relationships between epidemiological risk factors and MDD subtypes (Pasman et al, MedRxiv 2024).
Objective 2.3 reveal underlying brain cell types for each subtype
Additionally, our work on single-cell RNA-sequencing data from the human post-mortem brain samples provided new insights into the brain cell types implicated in different MDD subtypes, advancing our understanding of the biological mechanisms involved (Yao et al, MedRxiv 2024).
Work Package 3: Developing a Clinically Useful Prediction Algorithm for MDD Treatment Outcomes
In WP3, we will focus on developing machine learning-based prediction models to forecast long-term treatment outcomes in MDD patients. Currently most effort have been allocated to the WP1-2. However, we have started developing an analysis protocol on predicting treatment resistance within individuals with MDD.
Work Package 1: Dissecting MDD Subtypes in Relation to Long-Term Treatment Outcomes
This work package focuses on examining the differential treatment outcomes in clinically-informed MDD subtypes and identifying novel, data-driven subtypes using Nordic register/EMR data.
Objective 1.1. Assess and compare long-term treatment outcomes in clinically-informed subtypes
To successfully complete this objective, we have reviewed the literature and have had multiple rounds of discussions with clinical psychiatrists regarding important outcomes relevant to MDD. We have now completed analyses comparing these outcomes at the population level and within siblings. Observational studies are often subject to the confounding factors. By comparing the associations at these two levels help to draw conclusions that are closer to causal inference. The findings are clear that more severe subtypes (e.g. severe MDD) have much worse treatment outcomes than less severe ones (e.g. mild-moderate MDD). We are currently preparing a manuscript (Tong et al, in prep).
Objective 1.2. Identify data-driven subtypes that differ in treatment outcomes
We used comprehensive longitudinal data with over ten years of follow-up to assess treatment density, hospitalizations, recurrence, chronicity, diagnostic conversion, treatment resistance, disability, suicide, and premature mortality. A key achievement is the identification of new, data-driven MDD subtypes that offer more nuanced insights into treatment outcomes than the clinically informed subtypes. We further confirmed that these data-driven subtypes are replicable and genetically validated using family history measures. We are currently preparing a manuscript (Lu et al, in prep).
Work Package 2: Determining the Genetic and Epidemiological Causes Underlying MDD Subtypes
Objective 2.1. identify specific genetic risk factors for MDD subtypes
We are currently leading several large-scale genome-wide association studies (GWAS), identifying specific genetic loci associated with different MDD subtypes, including early-onset and late-onset MDD (Shorter et al, under review Nature Genetics), treatment-resistant MDD (Xiong et al, revision submitted to Molecular Psychiatry), and atypical MDD (Marshall et al, in prep).
Objective 2.2 test the causality of epidemiological risk factors in each subtype
The application of Mendelian randomization allowed us to identify causal relationships between epidemiological risk factors and MDD subtypes (Pasman et al, MedRxiv 2024).
Objective 2.3 reveal underlying brain cell types for each subtype
Additionally, our work on single-cell RNA-sequencing data from the human post-mortem brain samples provided new insights into the brain cell types implicated in different MDD subtypes, advancing our understanding of the biological mechanisms involved (Yao et al, MedRxiv 2024).
Work Package 3: Developing a Clinically Useful Prediction Algorithm for MDD Treatment Outcomes
In WP3, we will focus on developing machine learning-based prediction models to forecast long-term treatment outcomes in MDD patients. Currently most effort have been allocated to the WP1-2. However, we have started developing an analysis protocol on predicting treatment resistance within individuals with MDD.
At least two of the significant achievements above can be considered as advancing the field beyond the state-of-the-art.
First, the identification of novel MDD subtypes using advanced data science techniques represents a major leap forward. Previously it was not known why individuals with MDD might go on develop different outcome trajectories. From our findings, we could identify subgroups who did poorly because they had higher susceptibility to develop bipolar disorder, and similarly another group had higher susceptibility to develop schizophrenia or psychotic disorders. So, these results have challenged traditional clinical boundaries between MDD, bipolar disorder, and schizophrenia, and offers new opportunities for personalized treatment.
Second, our integration of genetic data with single-cell RNA sequencing to understand brain cell type involvement in MDD is a groundbreaking step in the search for the biological basis of psychiatric disorders. We showed that more brain cell-type enrichment in the more severe subtypes. This knowledge IS crucial for advancing experimental modelling and therapeutic development.
First, the identification of novel MDD subtypes using advanced data science techniques represents a major leap forward. Previously it was not known why individuals with MDD might go on develop different outcome trajectories. From our findings, we could identify subgroups who did poorly because they had higher susceptibility to develop bipolar disorder, and similarly another group had higher susceptibility to develop schizophrenia or psychotic disorders. So, these results have challenged traditional clinical boundaries between MDD, bipolar disorder, and schizophrenia, and offers new opportunities for personalized treatment.
Second, our integration of genetic data with single-cell RNA sequencing to understand brain cell type involvement in MDD is a groundbreaking step in the search for the biological basis of psychiatric disorders. We showed that more brain cell-type enrichment in the more severe subtypes. This knowledge IS crucial for advancing experimental modelling and therapeutic development.