Data-driven consideration of genetic disorders for global genomic newborn screening programs

Area of Science:

• Genomics
• Public Health Genetics
• Bioinformatics

Background:

• Newborn sequencing (NBSeq) is expanding genetic disorder screening.
• Significant variability exists in gene selection across international NBSeq programs.
• A systematic approach is needed to prioritize genes for NBSeq.

Purpose of the Study:

• To develop a systematic method for prioritizing genes for newborn sequencing.
• To identify key predictors influencing gene inclusion in NBSeq programs.
• To create a machine learning model for ranking genes for public health consideration.

Main Methods:

• Assembled a dataset of 25 characteristics for 4,390 genes across 27 NBSeq programs.
• Employed regression analysis to determine predictors of gene inclusion.
• Developed a machine learning model (boosted trees) using 13 predictors to rank genes.

Main Results:

• Gene inclusion varied widely across 27 NBSeq programs (134-4,299 genes).
• Only 74 genes (1.7%) were included in over 80% of programs.
• Key predictors for inclusion were: US Recommended Uniform Screening Panel listing (74.7% increase), natural history evidence (29.5%), and treatment efficacy (17.0%).
• The machine learning model achieved high accuracy (AUC=0.915, R²=84%) in predicting gene inclusion.

Conclusions:

• The developed machine learning model provides a ranked gene list for NBSeq.
• This model can adapt to evolving evidence and regional requirements.
• It facilitates more consistent and informed gene selection in NBSeq initiatives.