There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
Principal Investigator | |
Principal Investigator's Name: | Vera Afreixo |
Institution: | University of Aveiro |
Department: | Mathematics |
Country: | |
Proposed Analysis: | As a researcher at the University of Aveiro / Department of Mathematics, I intend to develop and validate procedures for analyzing large data, proposing more robust and stable methodologies. In particular, in the context of neurodegenerative diseases we intend to apply the procedures developed to model the occurrence of disease or severity of neurodegenerative disease (e.g. Alzheimer). |
Additional Investigators | |
Investigator's Name: | Ana Helena Tavares |
Proposed Analysis: | The main goal of the work that is being developed with ADNI Data is to find a consistent method that identifies a correlation between some Single Nucleotide Polymorphisms (SNPs) and Alzheimer’s disease in a structure with a huge number of potential predictor variables. To achieve this, penalized regression methods (Least Absolute Shrinkage and Selection Operator – LASSO; Elastic net) have been applied in a combined way with methods based on Akaike’s information criteria (AIC) to evaluate the importance of potential predictors. The penalized regression models allow us to reduce the data dimensionality. For each penalized regression method we generated 100 models, where the penalized parameter (λ) of each model was obtained through 10-fold-cross-validation. Within each method, we calculated the AIC for each model. Following that, we assigned to each model a weight, which was calculated through the formula bellow: wi=exp((-1/2)∆i)/sum(exp(-1/2∆i)) , where ∆i=AICi-AICmin, i = 1 ,...,100 This way, based on the weights assigned to each model, it was possible to obtain a weight for each SNP. We consider that important SNPs to explain the dependent variable weight at least 0.8. Literature references have been found about this threshold. The applied methods still need to be improved to get our main goal. All statistics analysis are being performed in R (version 4.1.0) |
Investigator's Name: | Miguel Pinheiro |
Proposed Analysis: | The main goal of the work that is being developed with ADNI Data is to find a consistent method that identifies a correlation between some Single Nucleotide Polymorphisms (SNPs) and Alzheimer’s disease in a structure with a huge number of potential predictor variables. To achieve this, penalized regression methods (Least Absolute Shrinkage and Selection Operator – LASSO; Elastic net) have been applied in a combined way with methods based on Akaike’s information criteria (AIC) to evaluate the importance of potential predictors. The penalized regression models allow us to reduce the data dimensionality. For each penalized regression method we generated 100 models, where the penalized parameter (λ) of each model was obtained through 10-fold-cross-validation. Within each method, we calculated the AIC for each model. Following that, we assigned to each model a weight, which was calculated through the formula bellow: wi=exp((-1/2)∆i)/sum(exp(-1/2∆i)) , where ∆i=AICi-AICmin, i = 1 ,...,100 This way, based on the weights assigned to each model, it was possible to obtain a weight for each SNP. We consider that important SNPs to explain the dependent variable weight at least 0.8. Literature references have been found about this threshold. The applied methods still need to be improved to get our main goal. All statistics analysis are being performed in R (version 4.1.0) |
Investigator's Name: | Leonor Rodrigues |
Proposed Analysis: | The main goal of the work that is being developed with ADNI Data is to find a consistent method that identifies a correlation between some Single Nucleotide Polymorphisms (SNPs) and Alzheimer’s disease in a structure with a huge number of potential predictor variables. To achieve this, penalized regression methods (Least Absolute Shrinkage and Selection Operator – LASSO; Elastic net) have been applied in a combined way with methods based on Akaike’s information criteria (AIC) to evaluate the importance of potential predictors. The penalized regression models allow us to reduce the data dimensionality. For each penalized regression method we generated 100 models, where the penalized parameter (λ) of each model was obtained through 10-fold-cross-validation. Within each method, we calculated the AIC for each model. Following that, we assigned to each model a weight, which was calculated through the formula bellow: wi=exp((-1/2)∆i)/sum(exp(-1/2∆i)) , where ∆i=AICi-AICmin, i = 1 ,...,100 This way, based on the weights assigned to each model, it was possible to obtain a weight for each SNP. We consider that important SNPs to explain the dependent variable weight at least 0.8. Literature references have been found about this threshold. The applied methods still need to be improved to get our main goal. All statistics analysis are being performed in R (version 4.1.0) |
Investigator's Name: | Gabriela Moura |
Proposed Analysis: | The main goal of this work was to find a stable and accurate procedure that identifies association between relevant single nucleotide polymorphisms and alzheimer’s disease in a structure where we have a huge number of SNPs |
Investigator's Name: | Vera Enes |
Proposed Analysis: | The main goal of this work was to find a stable and accurate procedure that identifies association between relevant single nucleotide polymorphisms and alzheimer’s disease in a structure where we have a huge number of SNPs |