Phenotypic Resistance Diversity Index (PRDI) - Conceptual Validation
1. Calculation
PRDI was calculated for both synthetical and biological samples as the total length of the branches of the smallest subtree connecting the observed ARGs to the root of the PhenoARGTree, following the Faith’s phylogenetic diversity formulation. This calculation was performed using the faith_pd () function from the scikit-bio library (v0.6.3).
2. Validation
Conceptual validation of the PRDI was carried out using synthetic samples to distinguish between redundant augmentation of genes sharing the same resistance phenotype and “True” functional diversification. (addition of genes carrying distinct phenotypes). Starting with an initial subset of 50 randomly selected ARGs with unique phenotypes, the total number of genes was increased in increments of +5, +10, +20, +40, +80, and +100%, either by adding genes with the same phenotype (case 1: redundancy) or genes with new phenotypes (case 2: diversification). The α-diversity indices, including richness and Shannon, were calculated using scikit-bio (v0.6.3). The PRDI difference between the two cases as a function of the percentage added was modeled using the ordinary least squares method with robust standard errors (HC3).
Mean curves ± standard error of richness, Shannon index, and PRDI (from top to bottom) as a function of the percentage of genes added (%), according to two scenarios: Case 1: redundancy (addition of genes with a similar phenotype) and Case 2: diversification (addition of genes with different phenotypes). Data from 30 synthetic samples per level. Richness and Shannon index increase in both scenarios, while PRDI is only sensitive to phenotypic diversification (case 2) ; ΔPRDI = PRDI (diversification) - PRDI (redundancy) (OLS, robust SE HC3): β ≈ 1.92 PRDI / +10% (95% CI [1.83–2.01], p = 3.3×10⁻⁸).