Deafness gene screening based on a multilevel cascaded BPNN model.
Xiao Liu1, Li Teng2, Wenqi Zuo3
1School of Microelectronics and Communication Engineering, Chongqing University, 174 Shapingba District, Chongqing, 400044, China. liuxiao@cqu.edu.cn.
BMC Bioinformatics
|February 21, 2023
Summary
This study introduces a machine learning model to efficiently predict genes linked to sudden sensorineural hearing loss. The developed cascaded backpropagation neural network (BPNN) model accurately identifies potential deafness-associated genes, accelerating genetic research.
Area of Science:
- Genetics
- Computational Biology
- Otolaryngology
Background:
- Sudden sensorineural hearing loss is a prevalent otolaryngology condition.
- Mutations in inherited deafness genes are strongly linked to this condition.
- Traditional gene identification methods are accurate but resource-intensive.
Purpose of the Study:
- To develop a computational method for predicting deafness-associated genes.
- To improve upon existing gene screening techniques using machine learning.
- To identify novel candidate genes for inherited deafness.
Main Methods:
- Utilized cascaded backpropagation neural networks (BPNNs) for gene prediction.
- Trained the model using known deafness-associated genes and random chromosomal genes.
- Validated predictive performance using a large-scale human genome analysis.
Main Results:
- The cascaded BPNN model demonstrated superior performance in screening deafness genes.
- Achieved a mean AUC (Area Under the Curve) greater than 0.98 in tests.
- Identified 20 highly suspected deafness-associated genes, with 3 confirmed in existing literature.
Conclusions:
- The machine learning approach effectively screens for potential deafness-associated genes.
- This computational method accelerates the discovery of genes linked to hearing loss.
- Predictions offer valuable insights for future genetic research in deafness.
