CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1594
Title: Differentiating various patient groups across parkinsonism spectrum via a new dantzig-selector-type screener Authors:  Frederick Kin Hing Phoa - Academia Sinica (Taiwan) [presenting]
Jing-Wen Huang - National Tsing-Hua University (Taiwan)
Yan-Han Lin - National Taiwan University (Taiwan)
Shau Ping Lin - National Taiwan University (Taiwan)
Yi-Tzang Tsai - National Taiwan University (Taiwan)
Ming-Che Kuo - National Taiwan University Hospital (Taiwan)
Koji Ueda - Japanese Foundation for Cancer Research (Japan)
Ruey-Meei Wu - National Taiwan University Hospital (Taiwan)
Abstract: A robust, efficient and easily accessible differential diagnosis (D/D) scheme of patient groups from the parkinsonism spectrum is presented. The scheme includes a newly proposed Biomedical Oriented Logistic Dantzig Selector (BOLD Selector) for identifying robust biomarkers for D/D from the main effect model of a supersaturated experiment with binary response. To test the scheme's robustness via a published lipidomic dataset, double cross-validation is implemented to assess the generalizability of the model and single out suitable tuning parameters of the BOLD Selector. A newly generated proteomic dataset is investigated by profiling plasma EV proteins by LC/MS-MS from Parkinson's disease without dementia (PDND), PD with mild cognitive impairment (PD-MCI), PD with dementia (PDD), multiple system atrophy (MSA) and healthy controls (HC), using a multi-stage binary tree that employs a BOLD Selector at each stage to derive the prediction formulas for D/D of various patient groups. Biomarker candidates are engaged in the lipid metabolic pathway relevant to alpha-synucleinopathy, the pathological hallmark for both PD and MSA, indicating the promise of the BOLD Selector. Not only can it identify robust biomarkers with pathophysiological significance, thus facilitating D/D, but it can also pave the way towards identifying disease-relevant targets.