Schedule

Lecture 1

Time: 12:00 - 12:30

Speaker: Rick Reijnders

Title: "Machine learning-based prediction of cognitive outcomes in de novo Parkinson's disease"

Abstract: 
Cognitive impairment is a debilitating symptom in Parkinson’s disease (PD). We aimed to establish an accurate multivariate machine learning (ML) model to predict cognitive outcome in newly diagnosed PD cases from the Parkinson’s Progression Markers Initiative (PPMI). Annual cognitive assessments over an eight-year time span were used to define two cognitive outcomes of i) cognitive impairment, and ii) dementia conversion. Selected baseline variables were organized into three subsets of clinical, biofluid and genetic/epigenetic measures and tested using four different ML algorithms. Irrespective of the ML algorithm used, the models consisting of the clinical variables performed best and showed a better prediction of cognitive impairment outcome over dementia conversion. We observed a marginal improvement in the prediction performance when clinical, biofluid and epigenetic/genetic variables were all included in one model. Several cerebrospinal fluid measures and an epigenetic marker showed high predictive weighting in multiple models when included alongside clinical variables.

Lecture 2

Time: 12:30 - 13:00

Speaker: Valentin Laroche

Title: "Methylomic-based molecular subtyping of Alzheimer’s Disease"

Abstract: 
Accumulating evidence shows that the heterogeneity and temporal complexity of Alzheimer’s Disease contribute to the lack of effective treatment. Advances in omics technologies along with development of efficient integrative analytical pipelines, provide a unique opportunity to investigate the drivers of heterogeneity at multiple molecular levels. In this study, we used DNA methylation data quantified in postmortem prefrontal cortex to identify the molecular subtypes of AD, and the distinct molecular features driving the disease heterogeneity.

We used DNA methylation data generated in postmortem prefrontal cortex in three independent biobanks of Brain for Dementia Research (BDR; n = 415), the Mount Sinai Brain Bank (MSBB; n = 250) and (ROSMAP; n = 710). Multiple unsupervised clustering methods including network-based, Bayesian and Ensemble approaches were applied on DNA methylation data quantified by the Illumina EPIC arrays in the discovery cohort (BDR). Following each clustering method, distinct features related to the identified subtypes were captured using classification models such as sparse partial least squares discriminant analysis and random forest. The most accurate models comprising the distinct features were used classifying the samples within the two other independent cohorts (MSBB, ROSMAP).

We identified multiple well-defined clusters of samples in the BDR cohort. We achieved good prediction values in the test data to classify the subtypes using the distinct methylation features related to the identified subtypes (AUC: 0.86 – 0.96). The best performing classification model was applied on two other independent cohorts and similarly classified samples across different cohorts were extensively characterized using relevant clinical and pathological information.