Figure 2.

Multivariate analysis of immunologic assessment across samples of spleen, mesenteric lymph node, inguinal lymph node, peritoneal cavity lavage fluid, amniotic fluid, and placenta. (A) Recursive feature reduction used in an ensemble machine-learning strategy to determine the number of top features needed to achieve robust (>90% accuracy) classification. (B) Top 4 immune features that allow for distinction between s. Normalized expression levels are depicted. (C) Principal component analysis based on the top 4 features that allowed for optimal classification. (D) Individual classification algorithms were run with the top 4 features of the ensemble ranking. The receiver operating curve of Ridge regression is shown. The same results were observed using Passive-Aggressive or Logistic regression. Additional receiver operating curves are depicted in Supplementary Figure S2. AB, treated with antibiotics; MLN, mesenteric lymph nodes.