Fig. 1: MPSE can ingest different CNLP tool outputs and use alternative data types.
From: MPSE identifies newborns for whole genome sequencing within 48 h of NICU admission
Panels A, B display MPSE precision rates of patients manually selected for WGS (Panel A) and diagnostic yield for the subset of cases diagnosed by WGS (Panel B) using different CNLP tools. A CLiX-trained MPSE model from the RCHSD cohort was applied to phenotype data from 1838 University of Utah NICU patients generated by five different CNLP tools. Panels C, D display precision and diagnostic yield using MPSE models trained on four alternative data types (diagnosis codes, lab tests, medications, and all orders), compared to the corresponding HPO-based (CLiX) model trained on the same Utah cohort. A solid black reference line in each panel represents the precision or diagnostic yield expected from a model that chooses candidates at random, while the black dashed line in the diagnostic yield graphs (Panels B, D) indicates the NeoSeq study’s 40% total diagnostic yield. Figure generated with R ggplot2 software.
