AlzGenPred - CatBoost-based gene classifier for predicting Alzheimer's disease using high-throughput sequencing data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting memory.
• Vast amounts of AD-associated genomic data exist, but gene involvement requires further elucidation.
• Identifying AD-associated genes is crucial for understanding disease mechanisms and developing therapies.

Purpose of the Study:

• To develop a machine learning-based tool, AlzGenPred, for accurate identification of Alzheimer's disease-associated genes.
• To evaluate the efficacy of different feature types (sequence-based vs. network-based) in predicting AD-associated genes.
• To provide a user-friendly platform for accelerating AD biomarker discovery.

Main Methods:

• Generated 13,504 features from eight sequence-encoding schemes and evaluated 16 machine learning algorithms.
• Compared performance of sequence-based and network-based features for AD gene identification.
• Developed AlzGenPred using network-based features with a CatBoost machine learning model.

Main Results:

• Network-based features significantly outperformed sequence-based features in distinguishing AD-associated genes.
• Sequence-based features alone failed to achieve high classification accuracy.
• AlzGenPred achieved 96.55% accuracy and 98.99% AUROC, validated on the AlzGene dataset and transcriptomics data.

Conclusions:

• AlzGenPred is a highly accurate and reliable tool for identifying potential Alzheimer's disease biomarkers.
• The study highlights the superiority of network-based features for AD gene prediction.
• AlzGenPred facilitates accelerated biomarker discovery and enhances understanding of Alzheimer's disease.