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: | Eric Lim |
| Institution: | Nanyang Technological University |
| Department: | Lee Kong Chian School of Medicine |
| Country: | |
| Proposed Analysis: | Novel methods of network modelling and personalised machine learning to understand and predict individual dementia using heterogeneous multi-omics data: Dementia is known as a notoriously heterogeneous disease (For example, there are sub-classifications of both typical and atypical AD). When presenting with typical behavioural and mood disorders, dementia can also be misdiagnosed as psychiatric disease. Finally, dementia can result from Alzheimer's disease (AD) and Parkinson's disease. Such heterogeneity makes dementia very hard to study. Indeed, without clear biomarkers, dementia diagnosis is very challenging. Molecular profiling via gene and protein analyses may help generate useful diagnostic biomarkers. However, developing such biomarkers in the presence of high heterogeneity within any study cohort is also in itself challenging. Advancing research into dementia profiling and biomarker development will require analytical techniques allowing for precise sub-categorization of the disease cohort into more homogenous groups or disease states. Towards this end, as dementia research enters the realm of big data, with large data pools from consortia such as the Alzheimer’s Disease Neuroimaging Initiative, or ADNI (http://adni.loni.usc.edu/). Using ADNI data, we propose a two-tiered integrative approach combining network modelling and transductive personalised modelling (TPM) towards tackling heterogeneity of multi-omics profiling. We will refer to this henceforth as TPMNET. In Aim 1, we use TPMNET to characterize networks useful for subpopulation prediction during the earliest possible stage (pre-dementia). This allows checks if network approaches are sensitive, and can track the disease’s progression with network-based biomarkers. Indeed, if such networks are also associated with functionally coherent mechanisms, it may lead towards a deepened understanding of the subpopulation. In Aim 2, we will demonstrate how TPMNET produces good accuracy with also good explainability against any traditional (global) AI/ML modelling effort. TPMNET is a semi-supervised AI/ML approach that discovers subpopulations, and restricts models only to these. TPMNET minimizes heterogeneity amongst samples, while identifying personalised molecular signatures for each sample itself. In Aim 3, we will correlate the decision rules inferred from the molecular component of ADNI, and connect this with its neuroimaging and clinical data components. Deep phenotyping, based on methods such as neuroimaging, can provide useful insight into dementia progression but can be inconvenient and highly expensive to administer. Biomarker development that is closely associated with neuroimaging and clinical indicators is useful. We think after overcoming functional and etiological heterogeneities in molecular data, discovering clinically-relevant biomarkers is more attainable. |
| Additional Investigators | |
| Investigator's Name: | Wilson Goh |
| Proposed Analysis: | Novel methods of network modelling and personalised machine learning to understand and predict individual dementia using heterogeneous multi-omics data: Dementia is known as a notoriously heterogeneous disease (For example, there are sub-classifications of both typical and atypical AD). When presenting with typical behavioural and mood disorders, dementia can also be misdiagnosed as psychiatric disease. Finally, dementia can result from Alzheimer's disease (AD) and Parkinson's disease. Such heterogeneity makes dementia very hard to study. Indeed, without clear biomarkers, dementia diagnosis is very challenging. Molecular profiling via gene and protein analyses may help generate useful diagnostic biomarkers. However, developing such biomarkers in the presence of high heterogeneity within any study cohort is also in itself challenging. Advancing research into dementia profiling and biomarker development will require analytical techniques allowing for precise sub-categorization of the disease cohort into more homogenous groups or disease states. Towards this end, as dementia research enters the realm of big data, with large data pools from consortia such as the Alzheimer’s Disease Neuroimaging Initiative, or ADNI (http://adni.loni.usc.edu/). Using ADNI data, we propose a two-tiered integrative approach combining network modelling and transductive personalised modelling (TPM) towards tackling heterogeneity of multi-omics profiling. We will refer to this henceforth as TPMNET. In Aim 1, we use TPMNET to characterize networks useful for subpopulation prediction during the earliest possible stage (pre-dementia). This allows checks if network approaches are sensitive, and can track the disease’s progression with network-based biomarkers. Indeed, if such networks are also associated with functionally coherent mechanisms, it may lead towards a deepened understanding of the subpopulation. In Aim 2, we will demonstrate how TPMNET produces good accuracy with also good explainability against any traditional (global) AI/ML modelling effort. TPMNET is a semi-supervised AI/ML approach that discovers subpopulations, and restricts models only to these. TPMNET minimizes heterogeneity amongst samples, while identifying personalised molecular signatures for each sample itself. In Aim 3, we will correlate the decision rules inferred from the molecular component of ADNI, and connect this with its neuroimaging and clinical data components. Deep phenotyping, based on methods such as neuroimaging, can provide useful insight into dementia progression but can be inconvenient and highly expensive to administer. Biomarker development that is closely associated with neuroimaging and clinical indicators is useful. We think after overcoming functional and etiological heterogeneities in molecular data, discovering clinically-relevant biomarkers is more attainable. |
| Investigator's Name: | Neamul Kabir |
| Proposed Analysis: | Novel methods of network modelling and personalised machine learning to understand and predict individual dementia using heterogeneous multi-omics data: Dementia is known as a notoriously heterogeneous disease (For example, there are sub-classifications of both typical and atypical AD). When presenting with typical behavioural and mood disorders, dementia can also be misdiagnosed as psychiatric disease. Finally, dementia can result from Alzheimer's disease (AD) and Parkinson's disease. Such heterogeneity makes dementia very hard to study. Indeed, without clear biomarkers, dementia diagnosis is very challenging. Molecular profiling via gene and protein analyses may help generate useful diagnostic biomarkers. However, developing such biomarkers in the presence of high heterogeneity within any study cohort is also in itself challenging. Advancing research into dementia profiling and biomarker development will require analytical techniques allowing for precise sub-categorization of the disease cohort into more homogenous groups or disease states. Towards this end, as dementia research enters the realm of big data, with large data pools from consortia such as the Alzheimer’s Disease Neuroimaging Initiative, or ADNI (http://adni.loni.usc.edu/). Using ADNI data, we propose a two-tiered integrative approach combining network modelling and transductive personalised modelling (TPM) towards tackling heterogeneity of multi-omics profiling. We will refer to this henceforth as TPMNET. In Aim 1, we use TPMNET to characterize networks useful for subpopulation prediction during the earliest possible stage (pre-dementia). This allows checks if network approaches are sensitive, and can track the disease’s progression with network-based biomarkers. Indeed, if such networks are also associated with functionally coherent mechanisms, it may lead towards a deepened understanding of the subpopulation. In Aim 2, we will demonstrate how TPMNET produces good accuracy with also good explainability against any traditional (global) AI/ML modelling effort. TPMNET is a semi-supervised AI/ML approach that discovers subpopulations, and restricts models only to these. TPMNET minimizes heterogeneity amongst samples, while identifying personalised molecular signatures for each sample itself. In Aim 3, we will correlate the decision rules inferred from the molecular component of ADNI, and connect this with its neuroimaging and clinical data components. Deep phenotyping, based on methods such as neuroimaging, can provide useful insight into dementia progression but can be inconvenient and highly expensive to administer. Biomarker development that is closely associated with neuroimaging and clinical indicators is useful. We think after overcoming functional and etiological heterogeneities in molecular data, discovering clinically-relevant biomarkers is more attainable. |
| Investigator's Name: | Wenhao Han |
| Proposed Analysis: | Novel methods of network modelling and personalised machine learning to understand and predict individual dementia using heterogeneous multi-omics data: Dementia is known as a notoriously heterogeneous disease (For example, there are sub-classifications of both typical and atypical AD). When presenting with typical behavioural and mood disorders, dementia can also be misdiagnosed as psychiatric disease. Finally, dementia can result from Alzheimer's disease (AD) and Parkinson's disease. Such heterogeneity makes dementia very hard to study. Indeed, without clear biomarkers, dementia diagnosis is very challenging. Molecular profiling via gene and protein analyses may help generate useful diagnostic biomarkers. However, developing such biomarkers in the presence of high heterogeneity within any study cohort is also in itself challenging. Advancing research into dementia profiling and biomarker development will require analytical techniques allowing for precise sub-categorization of the disease cohort into more homogenous groups or disease states. Towards this end, as dementia research enters the realm of big data, with large data pools from consortia such as the Alzheimer’s Disease Neuroimaging Initiative, or ADNI (http://adni.loni.usc.edu/). Using ADNI data, we propose a two-tiered integrative approach combining network modelling and transductive personalised modelling (TPM) towards tackling heterogeneity of multi-omics profiling. We will refer to this henceforth as TPMNET. In Aim 1, we use TPMNET to characterize networks useful for subpopulation prediction during the earliest possible stage (pre-dementia). This allows checks if network approaches are sensitive, and can track the disease’s progression with network-based biomarkers. Indeed, if such networks are also associated with functionally coherent mechanisms, it may lead towards a deepened understanding of the subpopulation. In Aim 2, we will demonstrate how TPMNET produces good accuracy with also good explainability against any traditional (global) AI/ML modelling effort. TPMNET is a semi-supervised AI/ML approach that discovers subpopulations, and restricts models only to these. TPMNET minimizes heterogeneity amongst samples, while identifying personalised molecular signatures for each sample itself. In Aim 3, we will correlate the decision rules inferred from the molecular component of ADNI, and connect this with its neuroimaging and clinical data components. Deep phenotyping, based on methods such as neuroimaging, can provide useful insight into dementia progression but can be inconvenient and highly expensive to administer. Biomarker development that is closely associated with neuroimaging and clinical indicators is useful. We think after overcoming functional and etiological heterogeneities in molecular data, discovering clinically-relevant biomarkers is more attainable. |
