Computational Research
DREAM BIG’s work with and development of modelling strategies (including artificial intelligence) to increase early identification of children at-risk for later mental health problems can be found here.
This page contains:
Towards Health Resilience in Various Environments (THRIVE)
The THRIVE initiative aims to develop a susceptibility index to improve health resilience and predictive accuracy for chronic diseases from perinatal to early postnatal stages
Generalized Random Forest (GRF)
Generalized Random Forests (GRFs) extend traditional random forests to estimate a variety of statistical quantities, including treatment effects and conditional quantiles.
Latent Environmental and Genetic InTeraction model (LEGIT)
LEGIT is an innovative approach to construct latent features within a regression model, facilitating intricate interactions in genetic and environmental studies.
Latest NEWS
| 2022/03/01 | Our latest paper published in frontiers in neuroscience. See here. |
| 2024/09/09 | Youth focus group starting recruitment Click here! |
Towards Health Resilience in Various Environments
The THRIVE initiative (Towards Health Resilience in Various Environments) is a pivotal research endeavor dedicated to enhancing health resilience from the perinatal to the early postnatal stages of life. By developing a comprehensive susceptibility index, THRIVE aims to improve the predictive accuracy of chronic diseases, identify modifiable factors, and inform effective preventive strategies. This initiative not only focuses on understanding the developmental origins of these conditions but also strives to refine and expand upon existing models, such as the Adverse Childhood Events Scales (ACES), to create more directly applicable measures of susceptibility across various clinical settings. Central to THRIVE’s mission is the goal to effectively communicate findings, engage with, and foster exchanges among key stakeholders to enhance understanding and implementation of health resilience strategies.
Generalized Random Forest (GRF)
Generalized Random Forests (GRFs) are an advanced extension of traditional random forests designed to estimate diverse statistical quantities beyond simple averages, such as treatment effects and conditional quantiles. By using a local estimation procedure and weighted combinations of observations, GRFs enhance the accuracy and interpretability of their predictions. This flexibility makes them particularly valuable in fields like personalized medicine, economics, and risk management, where understanding the variability of effects across different subpopulations is crucial. GRFs are non-parametric, allowing them to model complex relationships without assuming a specific functional form.
Latent Environmental and Genetic InTeraction model (LEGIT)
LEGIT revolutionizes the analysis of complex interactions in genetic and environmental data by creating latent features that represent weighted sums of multiple variables. This model enables more manageable and interpretable configurations of gene-environment interactions, reducing the complexity typically associated with such analyses. LEGIT’s efficiency in handling a wide array of variables makes it particularly suitable for studies with limited sample sizes. It employs an alternating optimization algorithm for training, which simplifies the non-linear complexities into manageable linear regressions. This approach allows for the straightforward estimation of model parameters, substantially reducing the number of required estimations compared to traditional models. LEGIT’s scalability and interpretability make it a valuable tool in the field, available for practical use through implementations in R and SAS. This model not only streamlines the analytical process but also ensures that smaller, correlated effects are not overlooked, providing a robust framework for understanding genetic and environmental interactions.