Abstract:Proteomics is playing an increasingly important role in determining Alzheimer's disease risk, with a particular goal in identifying markers of risk before the disease manifests itself clinically. The traditional analysis methods, however, have not been able to fully capitalize on the richness of genetic data. Machine learning analytic methods provide a novel automated, scalable and less subjective approach from which to identify unique targets. The study used proteomic and clinical data from a publicly available dataset that included 120 known signaling proteins measured from 259 archived plasma samples from individuals with Alzheimer's disease and non-demented controls (http://www.nature.com/nm/journal/v13/n11/suppinfo/nm1653_S1.html). We deployed a deep belief network (DBN) based method that uses an efficient training procedure that integrates unsupervised generative learning for feature discovery with a follow-up phase of supervised learning that fine tunes the features to optimize discrimination. The DBN method identified a subset of proteins (i.e. 20 proteins) that resulted in diagnostic accuracy greater than 90%, which was of almost equal predictability to computational methods that used all 120 proteins and was significantly better than classic machine learning methods, including Support Vector Machine, k-Nearest Neighbor and Back Propagation Neural Network. The resulting deep learning network indicated the weighted contribution of each protein to the final predictive model. We also found a strong relationship between ACRP30 and AD diagnosis. Our approach highlights the power of the deep learning model by addressing three application problems: disease detection, feature selection and identification of pathogenic factors. These results also found a strong link between ACRP30, an adipocyte-specific secreted protein involved in glucose metabolism, and risk for AD using a novel computational method and outperformed other machine learning approaches that were constrained by feature dimensionality and limited size data sets. This approach also simplified the measurement index required for diagnosis by identifying a much smaller subset of factors needed to achieve high diagnostic accuracy. These results also reinforced the link between metabolic risk factors and AD and suggest that apidonectin-linked pathways are a therapeutic drug target.

Identifying Alzheimer’s Disease-Related Proteins by LRRGD

A Novel Method for Identifying Alzheimer’s Disease-related Proteins

Identifying Protein Biomarkers in Blood for Alzheimer's Disease

Identifying Shared Diagnostic Genes and Mechanisms in Vascular Dementia and Alzheimer's Disease Via Bioinformatics and Machine Learning

Identification of feature genes and pathways for Alzheimer's disease via WGCNA and LASSO regression

An Alzheimers Disease Related Genes Identification Method Based on Multiple Classifier Integration

Identification of Diagnostic Biomarkers in Alzheimer's Disease by Integrated Bioinformatic Analysis and Machine Learning Strategies.

Identification of Blood Biomarkers for Alzheimer's Disease Through Computational Prediction and Experimental Validation.

DEEP LEARNING APPLICATION IN IDENTIFYING PROTEOMIC RISK MARKERS FOR ALZHEIMER’S DISEASE

Identifying the Role of Oligodendrocyte Genes in the Diagnosis of Alzheimer's Disease Through Machine Learning and Bioinformatics Analysis.

Identification of Blood-Based Glycolysis Gene Associated with Alzheimer's Disease by Integrated Bioinformatics Analysis

Identification of Alzheimer's Disease-Related Genes Based on Data Integration Method

An Improved Graph Laplacian Regularization Method for Identifying Biomarkers of Alzheimer's Disease

Identification of Novel Immune-Relevant Drug Target Genes for Alzheimer's Disease by Combining Ontology Inference with Network Analysis.

A Machine Learning Method for Identifying Critical Interactions Between Gene Pairs in Alzheimer's Disease Prediction.

Unveiling Immune-related Feature Genes for Alzheimer’s Disease Based on Machine Learning

Exploring the Key Genes and Identification of Potential Diagnosis Biomarkers in Alzheimer’s Disease Using Bioinformatics Analysis

Raman Spectroscopy Detection of Platelet for Alzheimer’s Disease with Predictive Probabilities

Integrative analysis reveals key lysosomal genes as potential therapeutic targets in Alzheimer's disease

Identification of Immune Cells and Key Genes associated with Alzheimer's Disease

Unveiling Immune-related feature genes for Alzheimer's disease based on machine learning