Newborn screening (NBS) is one of the most successful public health programs to identify newborns with different disorders that can be treated. There are more than 40 metabolic disorders on the Recommended Universal Screening Panel (RUSP) can be detected with metabolic data using mass spectrometry method from dried blood spots collected by heel stick shortly after birth. While successful in most respects, only a few biomarkers are used for each disorder in NBS with sensitivity favored over specificity, which leads to relatively high false positive rate in NBS. In order to reduce the number of false positives, we proposed to take a second-tier test for the newborns with screen-positive results in NBS.
Instead of using a single or a few biomarkers for each disorder, we tried to integrate all the metabolic data together with data mining methods to reduce the number of false positive samples. We can remove about half or more false positives in random forest and some other data mining methods. After removing some false positive samples with data mining, we proposed to use a validated multiplex NGS technology for sequence of 72 genes for inborn metabolic disorders.