Recursive Feature Elimination-based Biomarker Identification for Open Neural Tube Defects

Area of Science:

• Genetics and Bioinformatics
• Developmental Biology
• Artificial Intelligence in Medicine

Background:

• Open spina bifida (myelomeningocele) is a severe birth defect with unclear pathogenesis and limited prenatal treatment.
• Accurate diagnosis is crucial for managing myelomeningocele, necessitating advanced diagnostic tools.
• Machine learning (ML) offers potential for precision diagnosis in complex diseases.

Purpose of the Study:

• To identify key genes associated with open neural tube defects using an ML approach.
• To enhance the diagnostic accuracy of myelomeningocele through novel biomarkers.
• To provide insights into the molecular mechanisms underlying open neural tube defects.

Main Methods:

• Differential gene expression analysis was performed on amniotic fluid samples from datasets GSE4182 and GSE101141.
• Principal Component Analysis (PCA) was used for sample outlier detection.
• Recursive Feature Elimination (RFE), an ML technique, combined with differential gene expression analysis identified key genes. Validation was done using LR, DT, SVM, RF, and KNN classifiers with 5-fold cross-validation.

Main Results:

• Four key genes were identified: Growth Associated Protein 43 (GAP43), Glial fibrillary acidic protein (GFAP), Repetin (RPTN), and CD44.
• These genes are implicated in crucial neurological processes including axon growth, astrocyte differentiation, and neuroinflammation.
• The selected features demonstrated high accuracy in classifying diseased and healthy samples.

Conclusions:

• The identified genes (GAP43, GFAP, RPTN, CD44) serve as promising biomarkers for diagnosing and enabling early detection of open neural tube defects.
• These biomarkers can aid in evaluating disease progression and severity.
• Early detection through these biomarkers supports improved treatment and prevention strategies for open neural tube defects.