AI-Driven Rare Disease Identification: A Genomic Analysis Framework Using Deep Neural Networks
Neeraja Pullalarevu, Srikanth Chittipothu, Naga Venkatesh Gangabathula, Pruthvi Krishna Gutta, Purna chander Mashetty, Sathvik Reddy Chaganti
Abstract
Diagnostic rare disease is very difficult because of the lack of samples from patients, the complexity of genomic structure and high dimensional feature space. This brief paper attempts to argue a claims through playled devices AI-driven genomic analysis framework based on deep establishing neural networks (DNNs) to discover rare disease jointly with The Help of Next-Generation Sequencing data. The framework consists of a layer of an autoencoder-based autoencoder dimensionality reduction module, and a supervised DNN classifier to achieve the improvement of the efficiency of the learning and the reduction of the issue of over-fitting. The autoencoder, reduces on large genomic data to a low dimension latent space that is then taken by the DNN for disease accurate classification. The proposed model was evaluated on real genomic datasets using actual and it proved superior to the common classifiers of Support Vector Machines (SVM) and Random Forests in accuracy, precision, recall and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{F 1}$</tex>-score. Moreover, confusion matrix, and ROC analysis can also show the effectiveness of the model in robust and reliable detection even of the weakest disease-specific genomic signatures. This study shows that deep learning can be used to do extensible, accurate and data-efficient diagnosis of rare genetic diseases.