Genetic Risk Assessment of Nonsyndromic Cleft Lip with or without Cleft Palate by Linking Genetic Networks and Deep...

Area of Science:

• Genetics
• Bioinformatics
• Machine Learning

Background:

• Deep learning enhances disease risk classification.
• Population-based genetic studies face dimensionality challenges.
• Feature selection is crucial for accurate genetic risk assessment.

Purpose of the Study:

• To compare the predictive performance of GANNE with conventional methods for NSCL/P risk classification.
• To identify key genetic markers associated with NSCL/P risk.
• To evaluate the efficiency of GANNE in handling high-dimensional genetic data.

Main Methods:

• Case-control study in a Korean population.
• Application of genetic-algorithm-optimized neural networks ensemble (GANNE) for SNP selection and risk prediction.
• Comparison with polygenic risk score (PRS), random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and artificial neural network (ANN).
• Functional validation using gene ontology and protein-protein interaction (PPI) network analyses.

Main Results:

• GANNE demonstrated the highest predictive power, with a 10-SNP model achieving an AUC of 88.2%.
• GANNE improved AUC by 23% over PRS and 17% over ANN.
• The IRF6 gene was identified as a major hub gene, frequently selected by GA.
• Genes including RUNX2, MTHFR, PVRL1, TGFB3, and TBX22 significantly contributed to NSCL/P risk prediction.

Conclusions:

• GANNE is an efficient method for disease risk classification using a minimal set of optimal SNPs.
• The identified SNPs and genes provide insights into the genetic architecture of NSCL/P.
• Further validation is required to establish the clinical utility of GANNE for NSCL/P risk prediction.