Prediction of mild cognitive impairment using blood multi-omics data

Area of Science:

• Genomics
• Biomarkers
• Machine Learning

Background:

• Mild cognitive impairment (MCI) is an early stage of cognitive decline, bridging normal aging and Alzheimer's disease (AD).
• Accurate MCI diagnosis is challenging due to subtle symptoms, yet early detection is vital for intervention and preventing AD progression.
• Genomic data from blood offers a non-invasive method for MCI prediction, but existing machine learning models show limited performance.

Purpose of the Study:

• To develop and validate a machine learning model for predicting MCI using multi-omics blood data.
• To assess the predictive power of genomic structure data (CNVs) compared to gene expression data for MCI classification.
• To identify key genomic features and pathways associated with MCI.

Main Methods:

• Developed an XGBoost machine learning model utilizing gene expression and copy number variation (CNV) data from blood samples.
• Evaluated model performance using the Area Under the receiver operating characteristic Curve (AUC).
• Identified and analyzed important genomic features for MCI prediction.

Main Results:

• Achieved a high AUC of 0.9398 in classifying MCI patients from normal controls.
• Demonstrated that copy number variation (CNV) data is as informative as gene expression data for MCI prediction.
• Identified 149 significant genomic features, enriched in neurodegenerative disease-associated pathways like neuron development.

Conclusions:

• Blood-based multi-omics data, including CNVs, can effectively predict MCI.
• The study provides novel insights into the molecular characteristics of MCI.
• This approach holds promise for early MCI detection and intervention strategies to prevent Alzheimer's disease.