ABSTRACT
This research introduces a novel three-step approach employing deep learning algorithms to identify genetic variants associated with neurodegenerative diseases. In the first step, the entire genome is meticulously fragmented, and Convolutional Neural Networks (CNNs) are applied to isolate phenotype-associated fragments. The second step introduces a Sliding Association Evaluation (SAEv) method, calculating Phenotype Influence Rating (PIR) to identify Binary Nucleotide Polymorphisms (BNPs) associated with phenotypes. The final step employs comprehensive CNN analysis on all identified BNPs to construct a robust disease classification model. Performance evaluation metrics across 10 samples reveal consistently high sensitivity (0.85 to 0.93), specificity (0.86 to 0.92), accuracy, precision, and F1 scores (0.88 to 0.91). The results affirm the model's efficacy, striking a balance between sensitivity and specificity crucial for reliable disease identification. This three-step approach paves the way for precision medicine, offering a sophisticated means of unraveling the genetic complexities underlying neurodegenerative diseases.
