Skip to main content
. 2025 Sep 11;16:1630062. doi: 10.3389/fmicb.2025.1630062

Figure 4.

Diagram with six panels: (A) Donut chart depicting percentages of chemical compounds, with the largest portion as organic acids and derivatives at 29.56%. (B) PCA plot with REP and NEP groupings, showing statistically significant separation with p = 0.015. (C) OPLS-DA score plot showing differentiation between REP and NEP groups with R2 = 0.584. (D) Histogram of permutation test accuracy distribution peaking around 0.75, with an accuracy of 0.847. (E) Bar chart of VIP scores for various compounds, led by 2-hydroxyvaleric acid. (F) Bubble plot showing metabolic pathways' impact and significance, highlighting alanine, aspartate, and glutamate metabolism.

CSF metabolic profile analysis in REP and NEP. (A) Donut chart showing HMDB-based classification of all annotated metabolites. (B) PCA plot depicting metabolic variation between REP and NEP groups. (C) PLS-DA score plot illustrating distinct clustering of REP and NEP samples. (D) Validation of the PLS-DA classification model through permutation testing (n = 1000), with an empirical p-value of 0.044. (E) Important features with VIP > 1 identified by PLS-DA model. (F) Pathway analysis of metabolites with VIP > 1 in REP compared to NEP. CSF, cerebrospinal fluid; NEP, non-epilepsy; REP, refractory epilepsy; KEGG, Kyoto Encyclopedia of Genes and Genomes; HMDB, Human Metabolome Database; PCA, principal component analysis; PLS-DA, partial least-squares regression discriminant analysis; VIP, variable importance in projection.