Assoc. Prof. Yeşim Aydın Son will give a seminar on “Multi-step RF Modeling and Ensemble Approach for prioritization of risk SNVs Early and Differential Diagnosis of Late-Onset Alzheimer’s Disease (LOAD)” on 25 October at 13:40. The abstract of the talk and a short bio is shared below.

Bio:
Assoc. Prof. Yeşim Aydın Son is a medical scientist holding an M.D. and a Ph.D. in Genome Sciences and Technologies. She is a full-time faculty member and head of the Health Informatics Department in the Graduate School of Informatics, METU. Her research focus is the modeling of chronic and complex diseases, such as cancer and neurodegenerative diseases, based on integrated genomic and clinical data. Dr. Aydın Son integrates genome sciences and biomedical informatics interpreting high throughput molecular data through computational techniques and contributes to personalized medicine using data mining techniques, systems biology approaches, and molecular interactions.

Abstract:
LOAD is the most common type of dementia in the aging population, whose diagnosis is limited by clinical scales with high inter-application variability, costly imaging methods, and interventional tests. LOAD has a complex genetic etiology and the molecular mechanisms involved are still unclear. GWAS examines the statistical interactions of variants of individuals by univariate analysis. Machine learning algorithms can capture hidden, novel, and significant patterns by considering nonlinear interactions where multiple variants determine the risk. The optimized LOAD-RF-RF models were created with 74.0, 72.1%, and 85.1% accuracy rates for ADNI, NCRAD, and GenADA datasets, and LOAD-RF-RF models identified a total of 719 variants as highly associated with LOAD. For the meta-analysis, an ensemble scoring algorithm was developed that prioritizes consecutive, common genomic-located variants in three predictive models at gene and LD levels. The protein-coding variants prioritized were selected for experimental validation based on their relationship with LOAD-related biological pathways after network, PPI, and enrichment analysis. For a total of 32 variants, pyrosequencing primers are designed and optimized. Model performance analysis is done with a case-control group of 43 LOAD diagnosed and 38 healthy participants, where 12 variants classified the LOAD risk with 79% precision and 7 protective variants presented a classification performance with 76% precision. A total of 26 variants were also found to be distinctive in the European non-Finnish (NFE) population. (Founding: TUBİTAK 1003 216S468)