A Novel GNN-Bi-LSTM Framework with Attention Mechanism for Detecting Rare Genetic Disorders

The growing population in the world and inadequate health care guidelines have caused the onset of sporadic genetic disorders that need effective diagnostic and prognostic methods. However, the existing systems cannot handle the multisystem nature of genetic information resulting in delays of formal identification and therapeutic procedures.

The proposed study aims at proposing a more reliable model, GNN-Bi-LSTM-AM, that incorporates Graph Neural Networks (GNN), Bi-LSTM, and attention module to improve the process of identifying and diagnosing rare genetic disorders (RGDs). In this model, GNN is used to describe the genetic factors' associations with other factors, whereas Bi-LSTM models the health data over time. The attention mechanism puts emphasis on sig-nificant genetic markers, which improves predictive ability and minimizes interpretation challenges. The framework makes use of complex data such as maternal health history and parental genetic inheritance. In particular, the GNN learns the association between patients on the basis of genetic similarities, and Bi-LSTM computes temporal details of health information, and the attention module is applied to pay attention to important elements.

The results of the experiment show that the proposed GNN-Bi-LSTM-AM model predicts the accuracy of 98.5, precision of 96, recall of 96, and an F1-score of 96, which is much better than traditional machine learning methods. The framework has also good efficiency in that the processing time was 7.85 seconds.

The GNN-Bi-LSTM-AM framework suggested is a great step towards predictive analytics of rare genetic diseases. The model can be used to provide promising applications in early detection and clinical diagnosis by combining genetic relationships, sequential health data, and feature relevancy. Therefore, it is practical in enhancing treatment interventions and patient outcomes in genetic medicine.