Identifying Patients with Atrioventricular Septal Defect in Down Syndrome Populations by Using Self-Normalizing...

Area of Science:

• Computational biology
• Genetics
• Medical informatics

Background:

• Atrioventricular septal defect (AVSD) is a critical congenital heart disease (CHD) often associated with Down syndrome (DS).
• Understanding genetic variations between DS patients with and without AVSD is crucial for elucidating their complex relationship.

Purpose of the Study:

• To develop and evaluate a computational method for distinguishing DS patients with AVSD from those without.
• To identify optimal genetic features associated with AVSD in DS patients.

Main Methods:

• Utilized copy number probes on chromosome 21 for patient encoding.
• Employed Monte Carlo feature selection (MCFS) for feature ranking.
• Applied a two-stage incremental feature selection with self-normalizing neural networks (SNNs) for classification and optimal feature identification.

Main Results:

• Identified 2737 optimal genetic features using the SNN approach.
• Achieved a Matthew's correlation coefficient (MCC) of 0.748 with the optimal SNN classifier.
• Compared to random forest, which yielded an MCC of 0.582 with 132 features.

Conclusions:

• The proposed SNN-based computational method effectively distinguishes DS patients with AVSD based on genetic features.
• The identified optimal features provide insights into the genetic underpinnings of AVSD in the context of Down syndrome.
• Further analysis of key features supports their essential roles in the condition.