Fig. 4

Random Forest Regressor based relative feature importance for the top 35 features predicting subsequent infant growth (y-axis shows feature importance scores which do not have a specific unit). Key (alphabetical): AGP – Alpha- 1-acid Glycoprotein; CRP – C-reactive Protein; GLP2 – Glucagon-like peptide 2; HAZ - Height for Age z-score; HuEotaxin – Human Eotaxin; HuGCSF – Human Granulocyte-colony stimulating factor; HuIL4 – Human Interleukin-8; HuIL7 – Human Interleukin-7; HuIL8 – Human Interleukin-8; HuIL9 – Human Interleukin-9; HuILra – Human Interleukin-1 Receptor Antagonist; HuIP10 – Human Interferon gamma-induced protein 10;HuPDGFb – Human Platelet Derived Growth Factor Subunit B; HuRANTES – Human RANTES (CCL5; C-C Motif Chemokine Ligand 5); HuTNFa – Human Tumor necrosis factor-α; HuVEGF – Human Vascular Endothelial Growth Factor; LPSIgAOD – Lipopolysaccharide IgA Optical Density; LPSIgGOD – Lipopolysaccharide IgG Optical Density; MotherLiterate – Literacy Status of the Mother; MPO – Myeloperoxidase; NEO – Neopterin; Reg1Serum – Serum Regenerating Gene 1β. Note: ‘ln’ before a variable refers to natural logarithm, ‘_9mo’ after a variable refers to the biomarker being collected at 9 months of follow-up, variables without mention of a time frame were collected at birth